To begin searching the abstracts, please use the search feature above.
| 841 | Computer 66
|
Accelerated MRI Thermometry with Multi-Echo Multi-Slice GRE |
| Yuval Zur1, Itzhak Pinhas1, and Boaz Shapira1 | ||
1Insightec Ltd, Haifa, Israel |
||
An accelerated multi echo GRE sequence is presented for Focused Ultrasound Thermometry. An under sampled k space dataset is acquired during heating, to reduce scan time and enable multi-slice acquisition. Prior to heating, a fully sampled baseline image is acquired. Based on the baseline image, an iterative reconstruction algorithm fills-in the missing lines of each hot image. Accelerations of up to x15 with 16 echoes enable acquisition of 5 – 10 slices in 3.5 sec. Real-time reconstruction of 800 msec/5 slices was implemented on a FUS-MRI machine. |
| 842 | Computer 67
|
An untrained deep learning method with model-based regularization for reconstructing dynamic MR images from retrospectively accelerated data |
| Kalina P Slavkova1, Julie C DiCarlo2,3, Viraj Wadhwa4, Thomas E Yankeelov2,3,5,6,7, and Jonathan I Tamir2,4,6 | ||
1Department of Physics, The University of Texas at Austin, Austin, TX, United States, 2Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin, Austin, TX, United States, 3Livestrong Cancer Institutes, The University of Texas at Austin, Austin, TX, United States, 4Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, United States, 5Department of Biomedical Engineering, The University of Texas at Austin, Austin, TX, United States, 6Department of Diagnostic Medicine, The University of Texas at Austin, Austin, TX, United States, 7Department of Oncology, The University of Texas at Austin, Austin, TX, United States |
||
Acquiring high-resolution MRI data for tissue parameter mapping for quantitative imaging requires additional scan time. As a proof-of-principle, we evaluated the ability of the ConvDecoder architecture regularized with a physical model to reconstruct accelerated variable-flip angle MRI data of the brain for T1-mapping. The performance of our method was compared to non-regularized ConvDecoder, low rank reconstruction, and compressed sensing. Our results suggest that ConvDecoder with physics-based regularization may provide a stopping condition for training that is not dependent on the ground truth data while improving parameter mapping at higher accelerations. |
| 843 | Computer 68
|
Systematic Standardization of Deep Learning Based Accelerated MRI Reconstruction Pipelines |
| Beliz Gunel1,2, Arjun Divyang Desai1,2, Shreyas Vasanawala3, Akshay Chaudhari3,4, and John Pauly1 | ||
1Electrical Engineering, Stanford University, Stanford, CA, United States, 2Equal contribution, Stanford University, Stanford, CA, United States, 3Radiology, Stanford University, Stanford, CA, United States, 4Biomedical Data Science, Stanford University, Stanford, CA, United States |
||
Deep learning based accelerated MRI reconstruction pipelines have potential to enable higher acceleration factors compared to traditional methods with fast reconstruction times and improved image quality. Although there have been studies regarding model architecture, loss function, and k-space undersampling patterns; the effect of scanner parameters, variations in sensitivity map estimation, training data requirement, and engineering decisions during model optimization and evaluation on the reconstruction performance remain largely unexplored. We systematically study the impact of such and show that such data extent, re-processing, and metric computation impact performance to the same or at larger extents than new architectures and loss functions. |
| 844 | Computer 69
|
Deep Plug-and-Play multi-coil compressed sensing MRI with matched aliasing: the Denoising-P-VDAMP algorithm |
| Charles Millard1, Aaron Hess1, Boris Mailhe2, and Jared Tanner3 | ||
1Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, 2Siemens, Princeton, NJ, United States, 3Mathematical Institute, University of Oxford, Oxford, United Kingdom |
||
We present the Denoising Parallel Variable Density Approximate Message Passing (D-P-VDAMP) algorithm for multi-coil compressed sensing MRI with a learned prior. To our knowledge, D-P-VDAMP is the first Plug-and-Play method for multi-coil k-space data where the distribution of the training data's aliasing matches the actual distribution seen during reconstruction. We evaluate the performance of the proposed method on the fastMRI knee dataset and find substantial improvements in reconstruction quality compared with Plug-and-Play FISTA with the same network architecture in similar training and reconstruction time. |
| 845 | Computer 70
|
Parallel Greedy Learning for Accelerating Cardiac Cine MRI |
| Batu M Ozturkler1, Arda Sahiner1, Tolga Ergen1, Arjun D Desai1, Christopher M Sandino1, Shreyas Vasanawala2, John M Pauly1, Morteza Mardani1, and Mert Pilanci1 | ||
1Electrical Engineering, Stanford University, Stanford, CA, United States, 2Radiology, Stanford University, Stanford, CA, United States |
||
Deep unrolled models have recently shown state-of-the-art performance for reconstruction of dynamic MR images. However, training these networks via backpropagation is limited by intensive memory and compute requirements to calculate gradients and store intermediate activations per layer. Motivated by these challenges, we propose an alternative training method by greedily relaxing the training objective. Our approach splits the end-to-end network into decoupled network modules, and optimizes each module separately, avoiding the need to compute end-to-end gradients. We demonstrate that our method outperforms end-to-end backpropagation by 3.3 dB in PSNR and 0.025 in SSIM with the same memory footprint. |
| 846 | Computer 71
|
Deep generative MRI reconstruction for unsupervised Gibbs ringing correction |
| Lucilio Cordero-Grande1,2, Enrique Ortuño-Fisac1, Jonathan O'Muircheartaigh2,3, Jo Hajnal2, and María Jesús Ledesma-Carbayo1 | ||
1Biomedical Image Technologies, ETSI Telecomunicación, Universidad Politécnica de Madrid & CIBER-BBN, Madrid, Spain, 2Centre for the Developing Brain & Biomedical Engineering Department, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 3Department for Forensic and Neurodevelopmental Sciences, Institute of Psychiatry, Psychology and Neuroscience & MRC Centre for Neurodevelopmental Disorders, King's College London, London, United Kingdom |
||
MRI reconstruction is formulated as the retrieval of the parameters of a deep decoder network fitted to the observations by an image formation model including the truncation of high-frequency information. Solutions without Gibbs ringing at any prescribed image grid can be obtained naturally by the model without training or ad-hoc post-corrections. We present quantitative and visual results for spectral extrapolation of magnitude images at different scales in an in-silico experiment and a high resolution ex-vivo brain MRI scan. After minor modifications to deal with complex data, the architecture is applied to 2D parallel imaging showing promising visual results. |
| 847 | Computer 72
|
Evaluation of Neural Network Reconstruction of Undersampled Data using a Human Observer Model of Signal Detection |
| Joshua D Herman1, Marcus L Wong1, Sajan G Lingala2, and Angel R Pineda1 | ||
1Mathematics Department, Manhattan College, Riverdale, NY, United States, 2Roy J. Carver Department of Biomedical Engineering, University of Iowa, Iowa City, IA, United States |
||
We evaluated images from undersampled data using a U-Net with common metrics (SSIM and NRMSE) and with a model for human observer detection, the sparse difference-of-Gaussians (S-DOG). We also studied how the results vary when changing the loss function and training set size. We saw that the S-DOG model would choose an undersampling of 2X while SSIM and NRMSE would choose 3X. In previous work, human observers also chose a 2X acceleration. The S-DOG model led to the same conclusion as the human observers. This result was consistent with changes in training set size and loss function. |
| 848 | Computer 73
|
Deep Learning Reconstruction for FRONSAC |
| Zhehong Zhang1 and Gigi Galiana2 | ||
1Department of Biomedical Engineering, Yale University, New Haven, CT, United States, 2Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States |
||
This work is the first to apply deep learning to the reconstruction of images encoded with nonlinear gradients. We apply a model-based deep learning network (MoDL) to simulated FRONSAC images and compare these to a PSF-based matrix inversion as well as cg-SENSE. The MoDL based reconstruction did not significantly change the behavior of signal noise. However, results demonstrate that the model-based deep learning network can outperform traditional reconstruction methods at high undersampling factors. Simulations also suggests that the regularizing network has potential to correct for miscalibration in the nonlinear gradient trajectory. |
| 849 | Computer 74
|
Properties of 2D MR image reconstructions with deep neural networks at high acceleration rates |
| Matthew J. Muckley1, Tullie Murrell2, Alireza Radmanesh3, Florian Knoll4, Zhengnan Huang3, Anuroop Sriram5, Daniel K. Sodickson3, and Yvonne W. Lui3 | ||
1Facebook AI Research, New York, NY, United States, 2shaped.ai, New York, NY, United States, 3Radiology, NYU School of Medicine, New York, NY, United States, 4Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander University Erlangen-Nuremberg, Erlangen, Germany, 5Facebook AI Research, Menlo Park, CA, United States |
||
We assess the properties of deep neural network-reconstructed brain MR images in the high acceleration regime at factors up to 100. We have three contributions: 1) metrics on model performance from 2- to 100-fold accelerations, 2) a Monte Carlo procedure for scoring the quality of model reconstructions using only subsampled data, and 3) assessment of the acceleration effects on pathology in six cases. Our Monte Carlo procedure can estimate ground truth PSNR with coefficients of determination greater than 0.5 using only subsampled data. Our pathology results were stable in DNN reconstructions up to 8-fold acceleration. |
| 850 | Computer 75
|
Distortion Free Image Reconstruction using a Deep Neural Network for an MRI-Linac |
| Shanshan Shan1,2, Yang Gao3, Paul Liu1,2, Brendan Whelan1, Hongfu Sun3, Feng Liu3, Paul Keall1,2, and David Waddington1,2 | ||
1ACRF Image X Institute, University of Sydney, Sydney, Australia, 2Ingham Institute For Applied Medical Research, Sydney, Australia, 3School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Australia |
||
The advent of MRI-guided radiotherapy has elevated demand for high geometric fidelity imaging. However, gradient nonlinearity can cause image distortion, which limits the accuracy of radiotherapy. In this work, we develop a deep neural network, namely DFReconNet, to reconstruct distortion free images directly from raw k-space in real time. Experiments on simulated brain datasets and phantom images acquired from an MRI-Linac demonstrated the utility of the proposed method. |
| 851 | Computer 76
|
A deep learning based direct mapping method for EPI image reconstruction |
| Changheun Oh1, Sung Suk Oh2, Jun-Young Chung1, and Yeji Han1 | ||
1Gachon University, Incheon, Korea, Republic of, 2Daegu Gyeongbuk Medical Innovation Foundation, Daegu, Korea, Republic of |
||
Echo planar imaging (EPI) is one of the fastest MRI pulse sequences that can acquire whole k-space data during a single excitation. In EPI, partial acquisition can be used for reducing the TR and the effective TE can be further reduced as the portion of the acquired k-space is reduced. However, the quality of the image decreases drastically as the acquired k-space portion is reduced. In this study, we proposed a solution for image reconstruction from partially acquired EPI based on a deep-learning based direct mapping method, where the images are directly reconstructed from input k-space data. |
| 852 | Computer 77
|
Brain Tumor Segmentation Using 3D CMM-Net with Limited and Accessible MR Images |
| Yoonseok Choi1, Mohammed A. Al-masni1, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of |
||
In this work, we investigate the possibility of using a limited number of medical MR images for brain tumor segmentation. This can assist to perform the segmentation task using limited and accessible data in clinical practice. We observed that the FLAIR and T1CE are the most suitable images that maintain comparable segmentation performances similar to the case of using more data (i.e., T1, T2, FLAIR, and T1CE). To further improve the segmentation results, we augment the selected image pair and generate an additional fusion map using the Singular Value Decomposition (SVD). |
| 853 | Computer 78
|
Magnetic Resonance Spectroscopic Imaging Data Denoising by Manifold Learning: An Unsupervised Deep Learning Approach |
| Amirmohammad Shamaei1,2, Jana Starčuková1, Radim Kořínek1, and Zenon Starčuk Jr1 | ||
1Institute of Scientific Instruments of the Czech Academy of Sciences, Brno, Czech Republic, 2Department of Biomedical Engineering, Faculty of Electrical Engineering and Communication, Brno University of Technology, Brno, Czech Republic |
||
This work demonstrates an unsupervised deep-learning approach to MRS(I) data denoising, incorporating a non-linear model without relying on MR-theoretical physical prior knowledge. The implemented autoencoder learns the underlying low-dimensional manifold in high-dimensional data vectors representing MRS(I) voxels. The method, developed for denoising water-fat spectroscopic images, was tested against both numerical and acquired phantoms containing milk cream. The proposed method shows results comparable with other techniques in boosting SNR and has been found more robust in low concentration component denoising. Deep learning data denoising for MRSI might result in faster acquisition, vital for MRSI clinical application. |
| 854 | Computer 79
|
Explaining variation in DTI parameters with meningioma microscopy: A comparison between a neural network and an image-feature-based approach |
| Jan Brabec1, Magda Friedjungová2, Daniel Vašata2, Elisabet Englund3, Linda Knutsson1,4,5, Filip Szczepankiewicz6, Pia C Sundgren6, and Markus Nilsson6 | ||
1Medical Radiation Physics, Lund University, Lund, Sweden, 2Department of Applied Mathematics, Faculty of Information Technology, Czech Technical University, Prague, Czech Republic, 3Division of Oncology and Pathology, Department of Clinical Sciences, Lund University, Skåne University Hospital, Lund, Sweden, 4Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States, 5F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, 6Diagnostic Radiology, Clinical Sciences Lund, Lund University, Lund, Sweden |
||
Diffusion MRI reflects tissue microstructure and mean diffusivity (MD) and fractional anisotropy (FA) are often readily interpreted in terms of cellularity and cell anisotropy, respectively. Here, we investigated to which degree histological features accounts for their variations in fixed sections of meningioma tumors. Histological slices were quantified in terms of cellularity and cell anisotropy, or by a neural network. Results show that in some cases the majority of the variation can be attributed to cellularity whereas in others none. Similarly, in some samples only a minority of the variability is attributable to the variability in FA explained by neural network. |
| 855 | Computer 80
|
Real time visual feedback for reproducible head motion |
| Adam van Niekerk1, Johan Berglund2, Henric Ryden1,3, Tim Sprenger1,4, Ola Norbeck1,3, Enrico Avventi1,3, and Stefan Skare1,3 | ||
1Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden, 2Medical Physics, Uppsala University Hospital, Uppsala, Sweden, 3Neuroradiology, Karolinska University Hospital, Stockholm, Sweden, 4MR Applied Science Laboratory Europe, GE healthcare, Stockholm, Sweden |
||
The timing and magnitude of volunteer head motion has a significant impact on artefact level. It can therefore be difficult to reproduce motion correction results. Moreover, it is of interest to replicate patient motion retrospectively, in volunteers, to better understand the efficacy of a motion correction method. We attempt to solve this by stamping motion updates with a synchronisation “tag”. A rolling model is used to project the volunteer’s pose onto an MRI compatible display. The “tag” synchronised pose from a previous scan is then shown alongside the volunteer’s current pose, allowing them to track and hence reproduce the motion. |
| 856 | Computer 81
|
Real-time Motion Compensated ∆B0 Shimming with an AC/DC Shim Coil and Dual-Echo vNavs |
| Nicolas Arango1, Robert Frost2,3, Paul Wighton2, Jason Stockmann2,3, Ovidiu C Andronesi2,3, and Andre van der Kouwe2,3 | ||
1Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 2A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 3Department of Radiology, Harvard Medical School, Boston, MA, United States |
||
Changes in subject position move susceptibility interfaces and therefore ∆B0 field patterns in the brain. We apply a prospective real time (TR-to-TR) shim updating scheme using dual echo EPI volume navigators to correct motion-induced changes in ∆B0 fields to reduce distortion in 2D EPI. Shim fields were produced by a 32 channel AC/DC head shim array. TR-to-TR shimming reduced EPI distortion in all head positions in in vivo experiments. |
| 857 | Computer 82
|
Prospective Motion Correction in Kidney MRI Using Free Induction Decay Navigators |
| Cemre Ariyurek1, Tess E Wallace1, Tobias Kober2,3,4, Aziz Koçanaoğulları1, Simon K Warfield1, Sila Kurugol1, and Onur Afacan1 | ||
1Radiology, Boston Children's Hospital and Harvard Medical School, Boston, MA, United States, 2Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, 3Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 4LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland |
||
Abdominal MRI scans may require breath-holding to prevent image quality degradation, which can be challenging for patients. In this study, we used free induction decay navigators (FIDnavs) to correct motion prospectively. A short calibration scan was acquired prior to the prospectively corrected scan to create a linear motion model to translate the FIDnav signal into kidney displacement. Shallow, deep and continuous deep breathing scans were acquired with the proposed technique and reconstructed images were compared to those without motion correction. The proposed technique reduces blurriness and motion artifacts in the kidneys by correcting their position prospectively during the MRI acquisition. |
| 858 | Computer 83
|
Motion-corrected supine breast MRI with online reconstruction |
| Karyna Isaieva1, Pierre-André Vuissoz1, Lena Nohava1,2, Michael Obermann 2, Elmar Laistler2, Marc Fauvel3, Claire Dessale3, Nicolas Weber1, Jacques Felblinger1,3, and Freddy Odille1,3 | ||
1IADI, Université de Lorraine, INSERM, Nancy, France, 2High Field MR Center, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria, 3CIC-IT, CHRU de Nancy, INSERM, Nancy, France |
||
A standard breast MRI protocol implies prone position. Supine position has multiple advantages; however, it requires motion correction. In this work we present our results on retrospective nonrigid motion correction of supine breast MRI on 7 healthy volunteers. A respiratory belt was used to capture the chest movements. Images were reconstructed online and fully automatically, including physiological data recording and its transfer to the reconstruction server. It was shown that application of the selected motion correction method improves image sharpness on 32-47% which was cofirmed by visual observations. |
| 859 | Computer 84
|
Pilot Tone meets DISORDER: Improved data-driven motion-corrected brain MRI by leveraging Pilot Tone signal variations |
| Yannick Brackenier1,2,3, Thomas Wilkinson1,2,3, Lucilio Cordero-Grande1,2,4, Raphael Tomi-Tricot1,2,3,5, Philippa Bridgen1,2,3, Sharon Giles1,2,3, Enrico De Vita1,2,3, Shaihan J Malik1,2,3, and Joseph V Hajnal1,2,3 | ||
1Biomedical Engineering Department, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2Centre for the Developing Brain, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 3London Collaborative Ultra high field System (LoCUS), London, United Kingdom, 4Biomedical Image Technologies, ETSI Telecomunicación, Universidad Politécnica de Madrid and CIBER-BNN, Madrid, Spain, 5MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom |
||
DISORDER is an established retrospective data driven motion correction approach that uses optimised phase encoding, but otherwise unmodified 3D acquisitions. It is highly effective, but requires multiple lines of k-space to be grouped together for each motion state to be estimated, and this limits temporal resolution. At 7T, head motion can also be detected by “Pilot Tone”, which is an injected RF signal picked up by each coil in the head receiver array, but a calibration step is required. Here we combine DISORDER and Pilot Tone to achieve integrated calibration and show that improved motion correction can result. |
| 860 | Computer 85
|
Prospective 3D FatNav motion correction for 7T Terra |
| Krzysztof Klodowski1, Juraj Halabrin1, Iulius Dragonu2, and Chris Rodgers1 | ||
1Clinical Neurosciences, Wolfson Brain Imaging Centre, Cambridge, United Kingdom, 2Siemens Healthcare Limited, Frimley, United Kingdom |
||
Ultra-high field (7T) imaging can deliver images of brain structure and function at previously inaccessible spatial resolutions down to ~100um. However, subject motion, particularly for patients causes blurring which can lose some of the benefit of 7T imaging. The 3D FatNav approach to motion correction embeds short fat-excitation imaging modules in dead-time in an MRI pulse sequence. We are developing a prospective FatNav module for Siemens 7T Terra scanners. We present here a prospective FatNav module that embeds FatNav into a host pulse sequence, that separates FatNav and main image data during online reconstruction, that computes motion updates and that applies these in GRE imaging sequence in real time. |
| 861 | Computer 86
|
Comparison of two retrospective tracking techniques in presence of fast and slow motion. |
| Elisa Marchetto1,2, Kevin Murphy2,3, and Daniel Gallichan1,2 | ||
1School of Engineering, Cardiff University, Cardiff, United Kingdom, 2Cardiff University Brain Research Centre (CUBRIC), Cardiff University, Cardiff, United Kingdom, 3School of Physics and Astronomy, Cardiff University, Cardiff, United Kingdom |
||
We investigated the artifacts arising from different types of head motion during brain structural MR imaging and how well these artifacts can be compensated for using retrospective correction based on two different motion-tracking techniques: FatNavs and Tracoline systems. High image quality could be recovered in our slow-motion scenarios using both motion estimation techniques. Masking the non-rigid part of the neck during FatNav volumes registration led to higher image quality when large pitch-motion was present. The fast continuous motion scenario led to more severe image artifacts that could not be fully compensated by the retrospective motion correction techniques used. |
| 862 | Computer 87
|
On the limits of model-based reconstruction for retrospective motion correction in the presence of spin history effects |
| Aizada Nurdinova1, Stefan Ruschke1, and Dimitrios C. Karampinos1 | ||
1Department of Diagnostic and Interventional Radiology, School of Medicine, Technical University of Munich, Munich, Germany |
||
MRI reconstruction methods comprising model-based motion correction modules often rely on the assumption of absence or piecewise constant motion patterns during signal readout and on the insignificance of spin history effects. Furthermore, the success of such methods also depends on the characteristics and motion sensitivity of the employed sequence and measured tissue properties and may therefor limit the general utility of such approaches in a clinical context. The present study demonstrates the limitations of Fourier-model-based motion correction reconstruction schemes by subtle violations of the piecewise constant motion assumption using Bloch simulations a model-based motion-corrected reconstruction approach. |
| 863 | Computer 88
|
MoDGAN: Unsupervised rigid motion detection and correction with generative adversarial networks |
| Mu-Yul Park1, Seul Lee1, Kyu-Jin Jung1, Jisu Yun1, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic, Yonsei university, Seoul, Korea, Republic of |
||
Rigid motion artifacts, which are caused by subject’s motion during MRI acquisition, are a significant problem of image degradation. In this study, we propose a GAN-based method for unsupervised rigid motion detection and correction. The proposed method detects the motion-corrupted phase encoding lines using an anomaly score for motion detection. To reduce motion artifacts, we replace them with corresponding lines that generator yielded. We show that proposed method achieves high accuracy in detecting motion-corrupted line without any hardware or modification in sequence. Furthermore, the results of motion correction also showed noticeable performance. |
| 864 | Computer 89
|
Unsupervised Deep Learning using modified Cycle Generative Adversarial Network for rigid motion correction in pediatric brain MRI |
| Sewook Kim1, Seul Lee1, Jaeuk Yi1, Mohammed A. Al-Masni1, Sung-Min Gho2, Young Hun Choi3, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of, 2GE Healthcare, Seoul, Korea, Republic of, 3Department of Radiology, Seoul National University Hospital, Seoul, Korea, Republic of |
||
Motion artifact that appears due to patient movement in MRI scans is a major obstacle in pediatric imaging. Since acquiring paired motion-clean and motion-corrupted dataset is difficult, unsupervised deep learning such as Cycle Generative Adversarial Network (Cycle-GAN) has been used in motion correction tasks these days. In this study, we propose a rigid motion artifact reducing strategy with modified Cycle-GAN by replacing generator that converts motion-free to motion-corrupted domain with a motion simulator. Our proposed model outperforms in contrast preservation and reducing artifacts compared to the original Cycle-GAN as well as reduces training parameters. |
| 865 | Computer 90
|
Deep Learning-Based Respiratory Motion Correction in Free-Breathing Abdominal Diffusion-Weighted Imaging |
| Jinho Kim1,2,3, Fasil Gadjimuradov2,3, Thomas Benkert3, Thomas Vahle3, and Andreas Maier2 | ||
1Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander Universität Erlangen-Nürnberg, Erlangen, Germany, 2Pattern Recognition Lab, Department of Computer Science, Friedrich-Alexander Universität Erlangen-Nürnberg, Erlangen, Germany, 3MR Application Pre-development, Siemens Healthcare GmbH, Erlangen, Germany |
||
Since a single diffusion-weighted image can suffer from low SNR, multiple DWI repetitions can be averaged to improve SNR, which, however, can introduce blurring due to respiratory motion between different repetitions. Consequently, retrospective gating can be performed to overcome this problem. However, conventional retrospective gating has low SNR efficiency as it discards parts of the data and may result in certain slices to be missing for the desired motion state. This work proposes an efficient Deep Learning-based motion-correction method to improve conventional retrospective gating in free-breathing DWI, resulting in sharper images while maintaining image information from all acquired repetitions. |
| 866 | Computer 91
|
Optimized denoising and removal of partial-Fourier induced Gibbs ringing improves accuracy and robustness of DTI and DKI parameters |
| Jenny Chen1, Benjamin Ades-aron1, Hong-Hsi Lee2, Durga Kullakanda1, Saurabh Maithani1, Dmitry S. Novikov1, Jelle Veraart1, and Els Fieremans1 | ||
1Department of Radiology, NYU School of Medicine, New York, NY, United States, 2Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, United States |
||
Diffusion MRI (dMRI) is affected by noise and by artifacts such as Gibbs ringing and distortions. Using software phantoms as ground-truth, this study compares diffusion tensor imaging (DTI) and diffusional kurtosis imaging (DKI) parameter estimates to assess accuracy of various denoising and Gibbs removal methods, two key components of dMRI pipelines. An optimized Diffusion parameter EStImation with Gibbs and NoisE Removal (DESIGNER) pipeline is proposed, with non-local patch MP-PCA denoising and Removal of Partial-fourier induced Gibbs Ringing (RPG), that yields more accurate metrics, fewer outlier voxels in phantoms and more robust DTI/DKI maps in patient data. |
| 867 | Computer 92
|
Comparison of image and extracorporeal derived arterial input functions for DCE-MRI in mice using an multimodal cross-validation approach |
| Florian Gierse1, Juela Cufe1, Bastian Maus2, Katharina Kronenberg3, Michael Claesener4, Paulina Dorten1, Klaus Schäfers1, Sven Hermann1, Uwe Karst3, Cornelius Faber2, Michael Schäfers4, Florian Büther4, and Philipp Backhaus4 | ||
1European Institute for Molecular Imaging, University of Münster, Münster, Germany, 2Translational Research Imaging Center, University of Münster, Münster, Germany, 3Institute of Inorganic and Analytical Chemistry, University of Münster, Münster, Germany, 4Department of Nuclear Medicine, University Hospital Münster, Münster, Germany |
||
Quantitative precise measurements of the dynamic arterial blood concentration (AIF) are challenging in perfusion MRI. As solution an extracorporeal circulation approach for the AIF determination with two different vascular access ways is presented and cross validated with injection of radioactive contrast agent derivatives. Independent of the vascular access way, extracorporeal derived AIFs point towards a high quantitative precision, although dispersion must be considered. However, deconvolution methods based on single gamma variate functions end up in plausible AIFs. The use of novel rapid imaging techniques in connection with extracorporeal AIFs indicates high potential for precise quantitative dual PET/MRI Pharmacokinetic-Modeling in mice. |
| 868 | Computer 93
|
Mapping Rodent Brain Mechanical Properties In Vivo with Magnetic Resonance Elastography and Nonlinear Inversion |
| L. Tyler Williams1, Katrina A. Milbocker2, Seth R. Sullivan1, Ian F. Smith2, Eric Brengel2, Gillian LeBlanc2, Anna Y. Klintsova2, Matthew D. J. McGarry3, and Curtis L. Johnson1 | ||
1Biomedical Engineering, University of Delaware, Newark, DE, United States, 2Psychological & Brain Sciences, University of Delaware, Newark, DE, United States, 3Thayer School of Engineering, Dartmouth College, Hanover, NH, United States |
||
Developing preclinical MRE techniques that are comparable to human MRE is important for translational studies. This pilot study investigates the efficacy of the nonlinear inversion algorithm on rat brain MRE data. Whole-brain MRE scans were performed on female, Long-Evans rats using a custom MRE-EPI sequence and piezoelectric actuator with a resolution of 0.25 mm isotropic and with 600 Hz vibration. NLI parameters were adjusted for brain size and frequency. The resulting shear stiffness and damping ratio maps exhibited strong contrast between different anatomical regions. These results validate the use of NLI in preclinical MRE settings. |
| 869 | Computer 94
|
VERDICT MRI for tumor tissue characterization – separating viable tumor from necrosis |
| Mikael Montelius1, Lukas Lundholm1, Oscar Jalnefjord1,2, Eva Forssell-Aronsson1,2, and Maria Ljungberg1,2 | ||
1Medical radiation sciences, Institute of Clinical Sciences, University of Gothenburg, Gothenburg, Sweden, 2Medical Physics and biomedical engineering, Sahlgrenska University Hospital, Gothenburg, Sweden |
||
The biological meaning of MRI-based biomarkers of cancer can be studied using spatially matched histologically stained tumor sections. However, correlations are often weak or contradictory, possibly due to confounding necrosis that models fail to describe. In this work, we demonstrate the potential of using VERDICT MRI model parameters to separate necrosis from viable tumor tissue non-invasively. Radiation treated tumors demonstrating large necrotic and viable regions are imaged, excised and stained using immunohistochemical methods. Correlations between spatially matched VERDICT and immunohistochemical maps reveal that VERDICT clearly separates necrosis from viable tumor. |
| 944 | Computer 64
|
Gradual RAKI reconstruction merged with intermittent GRAPPA blurring moderation with limited scan-specific training samples |
| István Homolya1, Péter Kemenczky1, Peter Dawood2, Zoltán Vidnyánszky1, and Martin Blaimer3 | ||
1Brain Imaging Centre, Research Centre for Natural Sciences, Budapest, Hungary, 2Department of Physics, University of Würzburg, Würzburg, Germany, 3Magnetic Resonance and X-Ray Imaging Department, Development Center X-ray Technology EZRT, Fraunhofer Institute for Integrated Circuits IIS, Würzburg, Germany |
||
RAKI is a scan-specific k-space interpolation technique based on deep convolutional networks, which bears superior noise resilience compared to GRAPPA. However, RAKI may introduce severe blurring in image reconstruction due to reduced number of autocalibration signal lines at higher acceleration factors. We propose Gradual RAKI, which exhibits the benefit of mixing RAKI and GRAPPA in a preparatory block for data augmentation purposes prior to a conventional RAKI reconstruction. Data augmentation provides an effective way to create synthetic ACS lines out of 8 original ACS lines at 4-fold acceleration, while valuable features are retained from both RAKI and GRAPPA reconstruction methods. |
| 945 | Computer 65
|
MRI Regularization for Low-Count PET Image Recovery using Unsupervised Learning Without the Need of Training Set - A Multi-Tracer PET/MRI Study |
| Mario Serrano-Sosa1 and Chuan Huang1,2,3 | ||
1Biomedical Engineering, Stony Brook University, Stony Brook, NY, United States, 2Radiology, Stony Brook Medicine, Stony Brook, NY, United States, 3Psychiatry, Stony Brook Medicine, Stony Brook, NY, United States |
||
Unsupervised denoising is useful as it allows low-count PET image recovery without the need of paired training data(low-count/full-count). However, current unsupervised denoising models utilize Contrast-to-Noise Ratio as stopping criteria to optimize the image recovery process, which can be improved by considering structural information to maintain the integrity of gross anatomy. In this work, we proposed an MRI structural regularization loss function for low-count PET image recovery using an unsupervised learning model, which does not require paired training sets and demonstrated that the proposed method is superior in both qualitative and quantitative analyses for two radiotracers with very different physiological uptake. |
| 946 | Computer 66
|
Deep Learning based MR reconstruction for accelerated 3D-PREFUL ventilation assessment of post-COVID-19 patients from undersampled MR-images |
| Maximilian Zubke1,2, Filip Klimeš1,2, Andreas Voskrebenzev1,2, Marcel Gutberlet1,2, Agilo L Kern1,2, Robin A Müller1,2, Arnd J Obert1,2, Frank Wacker1,2, and Jens Vogel-Claussen1,2 | ||
1Institute of Diagnostic and Interventional Radiology, Hannover Medical School, Hannover, Germany, 2Biomedical Research in Endstage and Obstructive Lung Disease Hannover (BREATH), German Center for Lung Research (DZL), Hannover, Germany |
||
3D phase-resolved functional lung (3D-PREFUL) proton MRI enables a radiation-free and non-contrast-enhanced ventilation assessment of human lungs. However, generating high-quality images usually requires a long acquisition time. Acceleration can be achieved by undersampling k-space data, but the resulting violation of the Nyquist theorem leads to image artifacts. Deep learning (DL)-based reconstruction approaches are proposed as a solution for this dilemma. Two novel loss functions are introduced to create a deep learning based reconstruction, optimized for lung MRI. The feasibility of ventilation assessment, including ventilation defect identification, from 8x undersampled MR-images of post-COVID-19 patients, reconstructed by a neural network is demonstrated. |
| 947 | Computer 67
|
Coil-sketched unrolled networks for computationally-efficient deep MRI reconstruction |
| Julio A Oscanoa1, Batu Ozturkler2, Siddharth S Iyer3,4, Zhitao Li2,4, Christopher M Sandino2, Mert Pilanci2, Daniel B Ennis4, and Shreyas S Vasanawala4 | ||
1Department of Bioengineering, Stanford University, Stanford, CA, United States, 2Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 3Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Boston, MA, United States, 4Department of Radiology, Stanford University, Stanford, CA, United States |
||
Deep unrolled networks can outperform conventional compressed sensing reconstruction. However, training unrolled networks has intensive memory and computational requirements, and is limited by GPU-memory constraints. We propose to use our previously developed “coil-sketching” algorithm to lower the computational burden of the data consistency step. Our method reduced memory usage and training time by 18% and 15% respectively with virtually no penalty on reconstruction accuracy when compared to a state-of-the-art unrolled network. |
| 948 | Computer 68
|
A Cartilage-Specific Loss Function Improves Image Reconstruction Performance in Multiple Tissues of Clinical Interest |
| Aniket Tolpadi1,2, Francesco Calivà1, Misung Han1, Emma Bahroos1, Peder Larson1, Sharmila Majumdar1, and Valentina Pedoia1 | ||
1Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States, 2Bioengineering, University of California, Berkeley, Berkeley, CA, United States |
||
Most MRI image reconstruction algorithms are optimized for full-volume performance rather than specific tissues. Using a KIKI-Net style architecture and multi-component loss function in image space and k-space as baseline, we find a cartilage-specific loss function improves reconstruction performance at R=4 and R=8 in both cartilage and menisci. Thus, it may be possible to improve clinical utility of reconstruction pipelines across tissues of heightened clinical interest using a simple loss function weighting. Furthermore, full-volume standard reconstruction metrics worsened at R=4 and R=8 while tissue-specific metrics improved, calling into question whether these metrics are best for assessing reconstruction pipeline clinical utility. |
| 949 | Computer 69
|
IMPULSED model-based brain tumor microstructural parameter estimation with deep neural network |
| Jian Wu1, Taishan Kang2, Xinran Chen1, Lina Xu1, Jianzhong Lin2, Zhigang Wu3, Tianhe Yang2, Congbo Cai1, and Shuhui Cai1 | ||
1Xiamen University, Xiamen, China, 2Zhongshan Hospital Afflicated to Xiamen University, Xiamen, China, 3MSC Clinical & Technical Solutions, Philips Healthcare, Shenzhen, China |
||
This study assesses the feasibility of training a convolutional neural network (CNN) for IMPULSED (imaging microstructural parameters using limited spectrally edited diffusion) model fitting to diffusion-weighted (DW) data and evaluates its performance on a brain tumor (poorly differentiated adenocarcinoma) patient data directly acquired from clinical MR scanner. Comparisons were made with the results calculated from the non-linear least squares (NLLS) algorithm. More accurate and robust results were obtained by our CNN method, with processing speed several orders of magnitude faster than the reference method (from 5 min to 1 s). |
| 950 | Computer 70
|
New Denoising Neural Network for Diffusion Tensor Imaging Signal-to-Noise Ratio Enhancement |
| Po-Ting Chen1, Tzu-Yi Wang1, and Jyh-Horng Chen1 | ||
1National Taiwan University, Taipei, Taiwan |
||
本研究提出了一種新的神經網絡模型來提高 DTI 圖像的信噪比。從實驗結果來看，我們成功地將圖像的 SNR 提高了 3 倍，相當於減少了 9 倍的掃描時間。這將改善因信噪比不足而導致的 DTI 分析困難。 |
| 951 | Computer 71
|
Assessment of Instabilities of Conventional and Deep Learning Multi-Coil MRI Reconstruction at Multiple Acceleration Factors |
| Ruoxun Zi1, Patricia Johnson1,2, and Florian Knoll3 | ||
1Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, NY, United States, 2Center for Advanced Imaging Innovation and Research (CAI2R), NYU School of Medicine, New York, NY, United States, 3Department Artificial Intelligence in Biomedical Engineering, FAU Erlangen-Nuremberg, Erlangen, Germany |
||
Although deep learning has received much attention for accelerated MRI reconstruction, it shows instabilities to certain tiny perturbations resulting in substantial artifacts. There has been limited work comparing the stability of DL reconstruction with conventional reconstruction methods such as parallel imaging and compressed sensing. In this work, we investigate the instabilities of conventional methods and the Variational Network (VN) with different accelerations. Our results suggest that CG-SENSE with an optional regularization is also impacted by perturbations but shows less artifacts than the VN. Each reconstruction method becomes more vulnerable with higher acceleration and VN shows severe artifacts with 8-fold acceleration. |
| 952 | Computer 72
|
Improved Neural Network-Based Coil Compression |
| Elizabeth K. Cole1,2, Qingxi Meng1,2, Anishka Raina3, John M. Pauly2, and Shreyas S. Vasanawala4 | ||
1Equal Contribution, Stanford, CA, United States, 2Electrical Engineering, Stanford University, Stanford, CA, United States, 3The Harker School, San Jose, CA, United States, 4Rad/Pediatric Radiology, Stanford University, Stanford, CA, United States |
||
Coil compression is performed in magnetic resonance imaging (MRI) to enable smaller datasets and faster computation time. However, the traditional coil compression process is lengthy and lossy. In this work, we proposed a novel neural network-based coil compression method to achieve higher reconstruction accuracy and faster coil compression. Our method consistently achieved up to 1.5x lower NRMSE compared to SVD and GCC on the fastMRI knee dataset. The computational requirements of our method are practical, and inference runs 10 times faster than the traditional methods. |
| 953 | Computer 73
|
Hybrid image and k-space deep learning reconstruction exploiting spatio-temporal redundancies for 2D cardiac CINE |
| Siying Xu1, Patrick Krumm2, Andreas Lingg2, Haikun Qi3, Kerstin Hammernik4,5, and Thomas Küstner1 | ||
1Medical Image And Data Analysis (MIDAS.lab), Department of Interventional and Diagnostic Radiology, University Hospital of Tuebingen, Tuebingen, Germany, 2Department of Radiology, University Hospital of Tuebingen, Tuebingen, Germany, 3School of Biomedical Engineering, ShanghaiTech University, Shanghai, China, 4Lab for AI in Medicine, Technical University of Munich, Munich, Germany, 5Department of Computing, Imperial College London, London, United Kingdom |
||
Cardiac CINE MR imaging allows for accurate and reproducible measurement of cardiac function but requires long scanning times. Parallel imaging and compressed sensing (CS) have reduced the acquisition time, but the possible acceleration remained limited. Deep learning-based MR image reconstructions can further increase the acceleration rates with improved image quality. In this work, we propose a novel network for retrospectively undersampled 2D cardiac CINE that a) operates in image and k-space domain with interleaved architectures and b) utilizes spatio-temporal filters for multi-coil dynamic data. The proposed network outperforms CS and some other neural networks in both qualitative and quantitative evaluations. |
| 954 | Computer 74
|
Improvements of Image Quality of 1H and 129Xe MRI by Using an Advanced Acquisition and Reconstruction Method Coupled with Deep Learning |
| Samuel Perron1, Matthew S. Fox1,2, and Alexei Ouriadov1,2,3 | ||
1Physics and Astronomy, The University of Western Ontario, London, ON, Canada, 2Lawson Health Research Institute, London, ON, Canada, 3School of Biomedical Engineering, The University of Western Ontario, London, ON, Canada |
||
Accelerated MRI has the potential to significantly improve image quality without increasing costs, especially for low field strengths. A series of undersampled images are averaged for every unique permutation, and their SNR dependency is fitted to the Stretched-Exponential-Model. The proposed method was implemented in proactively undersampled phantom images at low field (0.074T) and in retroactively undersampled human lung images at high field (3T) using the FGRE pulse sequence; in all cases, SNR was significantly improved within the same scan duration compared to a fully-sampled image. Reconstruction artefacts were minimized or completely removed using a convolutional neural network. |
| 955 | Computer 75
|
Near Metal MRI with Compressed Sensing PETRA |
| Serhat Ilbey1, Michael Bock1, Matthias Jung2, Lukas Konstantinidis3, and Ali Özen1 | ||
1Dept. of Radiology, Medical Physics, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 2Department of Diagnostic and Interventional Radiology, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 3Department of Orthopaedics and Trauma Surgery, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany |
||
MRI near metallic implants require high bandwidth excitation and acquisition to minimize signal void and pile-up artefacts. In this work, compressed sensing PETRA (csPETRA) was modified to have intentionally a longer TE, so that the phase-encoding part of the k-space is extended. To prevent extremely long scan times, SPI points in csPETRA sequence are pseudo-randomly undersampled to reduce the scan time below 6 minutes. csPETRA reduced artifacts near metal significantly while preserving the anatomical details. Isotropic 3D MRI of two volunteers with mouth and knee prostheses was performed with csPETRA and the artifacts are significantly reduced compared to T1w-WARP sequence. |
| 956 | Computer 76
|
Gibbs-ringing Removal through Anti-aliased Deep Priors |
| Jaeuk Yi1, Chuanjiang Cui1, Kyu-Jin Jung1, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of |
||
Inspired by recent usage of Deep Image Prior (DIP) in the field of MRI that utilizes a powerful low-level image prior from a neural network architecture itself without any training dataset, we conduct k-space extrapolation using the deep prior for Gibbs-ringing removal in order to build general Gibbs-ringing correction algorithm without dependency on the dataset. We further improved the existing deep prior with the addition of anti-aliasing layers. The proposed deep prior method outperformed conventional non-learning methods quantitively and qualitatively in numerical simulations and in-vivo data. |
| 957 | Computer 77
|
Imaging near metal at 0.55 T using gradient-echo based sequences: Feasibility and Opportunities |
| Kübra Keskin1, Brian A. Hargreaves2,3,4, and Krishna S. Nayak1,5 | ||
1Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, 2Radiology, Stanford University, Stanford, CA, United States, 3Electrical Engineering, Stanford University, Stanford, CA, United States, 4Bioengineering, Stanford University, Stanford, CA, United States, 5Biomedical Engineering, University of Southern California, Los Angeles, CA, United States |
||
The large magnetic susceptibility difference between metallic implants and surrounding tissues causes severe MRI artifacts that scale with the B0 field strength. At conventional field strengths, spin-echo based multispectral imaging is used to mitigate this artifact and produce diagnostic images. At lower field strengths, such as 0.55 T, other strategies may be feasible. Here, we investigate gradient-echo based sequences for imaging near metal at 0.55 T, which provide high SNR efficiency and different contrast. |
| 958 | Computer 78
|
Low Rank Off-resonance Correction for Double Half-Echo k-Space Acquisitions |
| Mark Bydder1, Fadil Ali1, Vahid Ghodrati1, Chencai Wang1, Akifumi Hagiwara1, and Ben Ellingson1 | ||
1UCLA, Los Angeles, CA, United States |
||
The present study describes a model-based approach for correcting off-resonance in the context of short echo-time imaging using a double half-echo k-space acquisition. This technique employs center-out readouts in forward and reverse directions and is sensitive to off-resonance. The correction is simple and results in a marked reduction in blurring for a negligble cost in pre-processing. |
| 959 | Computer 79
|
A Deep Forward-Distortion Model for Unsupervised Correction of Susceptibility Artifacts in EPI |
| Abdallah Zaid Alkilani1,2, Tolga Çukur1,2,3, and Emine Ulku Saritas1,2,3 | ||
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, 2National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara, Turkey, 3Neuroscience Graduate Program, Bilkent University, Ankara, Turkey |
||
Echo planar imaging (EPI) requires the correction of susceptibility artifacts for further quantitative analyses. Images acquired in reversed phase-encode (PE) directions are typically used to estimate the susceptibility-induced field from EPI data. In this work, an unsupervised deep Forward-Distortion Network (FD-Net) is proposed for the correction of susceptibility artifacts in EPI: The field and underlying anatomically-correct image are predicted, subject to the constraint that forward-distortion of this image with the field explains the input warped images. This approach provides rapid correction of susceptibility artifacts, with superior performance over deep learning methods that unwarp input images based on a predicted field. |
| 960 | Computer 80
|
Inline artifact identification using segmented acquisitions and deep learning |
| Keerthi Sravan Ravi1,2, Marina Manso Jimeno1,2, John Thomas Vaughan Jr.2, and Sairam Geethanath2 | ||
1Department of Biomedical Engineering, Columbia University, New York, NY, United States, 2Columbia Magnetic Resonance Research Center, New York, NY, United States |
||
MR artifacts degrade image quality and affect diagnosis, requiring thorough examination by the MR technician, and reacquisition in some cases. We employ a combination of segmented acquisitions and a deep learning tool (ArtifactID) to perform more frequent updates to image quality during acquisition. ArtifactID identified wrap-around, Gibbs ringing and motion artifacts, with a mean accuracy of 99.43%. The segmented acquisitions for rescans resulted in a 12.98% time gain over the full-FOV sequence. In addition, ArtifactID alleviates burden on the MR technician via automatic artifact identification, saving image quality evaluation time and augmenting expertise. |
| 961 | Computer 81
|
Metal Artifact Corrected and Chemical Shift Encoded MRI |
| Collin J Buelo1,2, Timothy J. Colgan1, and Diego Hernando1,2 | ||
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, 2Medical Physics, University of Wisconsin-Madison, Madison, WI, United States |
||
Fat-suppressed MRI near metal remains a challenging technical problem. Chemical shift encoded (CSE) water/fat separation enables excellent SNR efficiency compared to short tau inversion recovery (STIR), but the large B0 field offsets and gradients complicate conventional CSE methods near metal. To address these challenges, we propose metal artifact correction and chemical shift encoding using optimized echo spacings, and with deep learning processing for reconstruction of water/fat separated imaging near metal. This method is able to correctly separate water and fat closer to a metallic surgical breast clip than conventional methods, without the drawbacks associated with STIR imaging. |
| 962 | Computer 82
|
Improved Detection and Localization of Artifacts using CutArt Augmentation |
| Ben A Duffy1, Greg Zaharchuk2, Enhao Gong1, and Keshav Datta1 | ||
1Subtle Medical Inc., Menlo Park, CA, United States, 2Stanford University, Stanford, CA, United States |
||
With the aim of improving the performance of an automated quality control system, we propose to use a data augmentation technique based on cropped patches of simulated artifacts (CutArt) instead of artifacts that are distributed across the entire image. This has the advantage of improving the artifact localization and quality control classification performance, as assessed by experiments on simulated as well as real artifact affected data. Localization experiments suggested that the CutArt model learns to focus on the tissue of interest instead of the image background. |
| 963 | Computer 83
|
Multi-domain Motion Correction Network for Turbo Spin-echo MRI |
| Jongyeon Lee1, Wonil Lee1, Namho Jeong1, and HyunWook Park1 | ||
1Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of |
||
Motion correction in MRI has been successfully performed by adopting deep learning techniques to correct motion artifacts with data-driven algorithms. However, many of these studies have narrowly utilized MR physics. In addition to the efficient motion correction method proposed for multi-contrast MRI, we expand our previous method to utilize data in k-space domain for a single-contrast MRI. Using the motion simulation based on MR physics, we propose a multi-domain motion correction method in both k-space and image domains. Our proposed method finely reduces motion artifacts using the multi-domain network. |
| 964 | Computer 84
|
Time-segmented and contrast-neutral motion correction using an end-to-end deep learning approach |
| Refaat E Gabr1, Masoud Edalati2, Lingzhi E Hu2, Adam Chandler2, Weiguo Zhang2, and Ponnada A Narayana3 | ||
1Diagnostic and Interventional Imaging, University of Texas Health Science Center at Houston, Houston, TX, United States, 2United Imaging Healthcare, Houston, TX, United States, 3University of Texas Health Science Center at Houston, Houston, TX, United States |
||
We developed an end-to-end deep learning-based motion correction technique that utilized the time-segmented nature of MRI data acquisition. Results of computer simulations and in vivo studies show the network to be highly effective in correcting motion artifacts. |
| 965 | Computer 85
|
Characterisation and Removal of Slice Dependent Artifacts in EPI Phase Images |
| Jannette Nassar1, Oliver Kiersnowski1, Patrick Fuchs1, and Karin Shmueli1 | ||
1Medical Physics and Biomedical Engineering, UCL, London, United Kingdom |
||
Although EPI phase images are useful (e.g. for Quantitative Susceptibility Mapping), they often contain phase inconsistencies in the slice-select direction which persist and can degrade QSM results. Here, we analysed three EPI datasets in healthy volunteers to characterise these phase inconsistencies and understand whether they occur or interact with interleaved or sequential slice acquisition order. We characterised a ~2Hz cryogen pump artifact in sequential data and slice-to-slice phase jumps in interleaved data. We modified a previously proposed QSM processing pipeline, including 2D (VSHARP) and 3D (PDF) background field removal that removed all through-slice artifacts observed. |
| 966 | Computer 86
|
Predicting changes in B0 field due to patient movement dynamically for MRI via Deep Learning |
| Stanislav Motyka1, Paul Weiser2, Beata Bachrata1, Lukas Hingerl1, Bernhard Strasser1, Dario Goranovic1, Gilbert Hangel1,3, Eva Niess1, Maxim Zaitsev4, Simon Robinson1, Georg Langs2, Siegfried Trattning1,5, and Wolfgang Bogner1 | ||
1High Field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 2Computational Imaging Research Lab, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 3Department of Neurosurgery, Medical University of Vienna, Vienna, Austria, 4High Field MR Center, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria, 5Christian Doppler Laboratory for Clinical Molecular MR Imaging, Medical University of Vienna, Vienna, Austria |
||
We proposed a U-net-based neural network to predict changes in B0 field due to patient movement. The network was trained and tested on 7T in vivo volunteer data. The method was compared against two other approaches. The estimated B0 maps were compared against the ground truth. The results suggest that prediction of B0 maps is feasible and by combination with external tracking a considerable improvement in data quality of B0-sensitive MRI methods could be achieved. |
| 967 | Computer 87
|
Creating a multi-centre harmonised surface-based MRI dataset for the Multi-centre Epilepsy Lesion Detection Project |
| Mathilde Ripart1, Hannah Spitzer2, Russell Shinohara3, MELD Consortium4, Torsten Baldeweg1, Sophie Adler1, and Konrad Wagstyl5 | ||
1Great Ormond Street Institute for Child Health, UCL, London, United Kingdom, 2Institute of Computational Biology, Helmholtz Zentrum München, Munich, Germany, 3Penn Statistics in Imaging and Visualization Center, University of Pennsylvania, Philadelphia, PA, United States, 4UCL, London, United Kingdom, 5Wellcome Centre for Human Neuroimaging, UCL, London, United Kingdom |
||
The Multi-centre Epilepsy Lesion Detection (MELD) project presents a methodology to harmonise a large heterogenous cohort of surface-based MRI data. Structural features were extracted from T1w and FLAIR images and pre-processed to reduce systematic site, scanner, and age-specific biases. The harmonised dataset enabled the characterisation of subtle radiological markers of focal cortical dysplasia (FCD), a cortical abnormality causing drug-resistant epilepsy. Machine-learning algorithms trained on the harmonised dataset improved the classification of FCD histopathologies. With open-source protocols and code, the MELD preprocessing pipeline offers a reproducible method to prepare large heterogeneous datasets for statistical analysis and machine-learning tasks. |
| 968 | Computer 88
|
Female patients with atopic dermatitis have reduced functional connectivity in the precuneus and inferior parietal lobe |
| Tie-Qiang Li1, Klas Nordlind2, Elvar Theodorsson3, and Tomas Jonsson4 | ||
1Karolinska Institutet, Stockholm, Sweden, 2Division of Dermatology and Venereology, Karolinska Institutet, Stockholm, Sweden, 3Department of Biomedical and Clinical Science, Linköping University, Linköping, Sweden, 4CLINTEC, Karolinska Institutet, Stockholm, Sweden |
||
A cohort of 32 atopic dermatitis (AD) patients and 32 matched healthy controls (HC) were recruited to study the neurological abnormalities which might be associated with this pruritic skin condition. The main findings are 1) reduced general functional connectivity in the precuneus, inferior parietal lobe and visual cortex was detected in females AD patients; 2) Severity of the symptoms was negatively correlated with functional connectivity in inferior parietal lobe, while salivary cortisol was positively associated. Resting-state fMRI can provide useful insight into the altered neurophysiology and clinically relevant assessment for the AD disorder. |
| 969 | Computer 89
|
Enzymatic activity monitoring through Dynamic Nuclear Polarization in Earth magnetic field |
| Dahmane Boudries1, Jean-Michel Franconi1, Sylvain R.A Marque2, Philippe Massot1, Philippe Mellet3, Elodie Parzy1, and Eric Thiaudiere1 | ||
1CNRS, Bordeaux, France, 2CNRS, Marseille, France, 3INSERM, Bordeaux, France |
||
Earth-field MRI can provide new contrasts leading to the observation of pathologies at the biochemical level. However detection sensitivity is poor at low-field. In a preliminary spectroscopic approach, it is proposed here to detect protease-driven hydrolysis of a nitroxide probe thanks to electron-nucleus Overhauser enhancement in a homemade double resonance system. The nitroxide probe is a six-line nitroxide whose lines are shifted according to its substrate/product state. The Overhauser enhancement frequency dependence was in agreement with theoretical calculations. Enzymatic conversion of the nitroxide substrate was observed which opens the way for the design of new low-field DNP-MRI systems. |
| 970 | Computer 90
|
The Effect of Label Crossing the Blood-CSF barrier on Partial Volume Correction: source of error or opportunity for quantification? |
| Leonie Petitclerc1,2,3, Lydiane Hirschler1,3, Iris Asllani4,5, and Matthias J.P. van Osch1,2,3 | ||
1C.J. Gorter Center for High Field MRI, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, 2Leiden Institute for Brain and Cognition (LIBC), Leiden, Netherlands, 3Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, 4Clinical Imaging Science Center, University of Sussex, Sussex, United Kingdom, 5Biomedical Engineering, Rochester Institute of Technology, Rochester, NY, United States |
||
Partial volume correction in ASL typically neglects CSF signal, however we have recently shown that some labeled signal crosses the blood-CSF barrier at a measurable level. Here we show the effect of including CSF in PVC, in simulated and in-vivo data. CSF-PVC reduced error on GM perfusion by ~10% in simulated data, with higher error when more CSF is present (including at longer LD and PLD). The difference between CSF-PVC and non-CSF-PVC GM signal in vivo was also ~10%. Comparing PVC signal in CSF to long-TE ASL signal suggests that BCSFB characterization may be possible without the acquisition of ultra-long-TE. |
| 971 | Computer 91
|
The relationship between structural asymmetry and shear stiffness asymmetry in the human brain |
| Lily Xiang1, Lucy Victoria Hiscox2, Curtis L. Johnson3, and Neil Roberts1 | ||
1University of Edinburgh, Edinburgh, United Kingdom, 2University of Bath, Bath, United Kingdom, 3University of Delaware, Newark, DE, United States |
||
On average the human brain has a global structural asymmetry comprising a counter-clockwise twist in the transverse plane and which is referred to as the cerebral torque. We have investigated whether there is a relationship between the magnitude of the torque and measures of brain shear stiffness obtained using Magnetic Resonance Elastography (MRE) in a group of Older Adults (AD) and a group of patients with Alzheimers DIsease (AD). |
| 972 | Computer 92
|
Quantifiable study of magnetic resonance super resolution reconstruction in Placenta Accreta Spectrum using Image Quality Metrics |
| Joanna Chappell1, Nada Mufti1,2, Patrick O'Brien3, George Attilakos3, Magda Sokolska4, Priya Narayanan2, Rosalind Aughwane2,3, Giles Kendall2,3, David Atkinson5, Sebastien Ourselin 1, Anna L David2,3,6, and Andrew Melbourne1,4 | ||
1School of Biomedical Engineering and Imaging Sciences (BMEIS), Kings College London, London, United Kingdom, 2Elizabeth Garrett Anderson Institute for Women's Health, University College London, London, United Kingdom, 3Women's Health Division, University College London, London, United Kingdom, 4Depart. Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 5Centre for Medical Imaging, University College London, London, United Kingdom, 6WEISS (Wellcome /EPSRC Centre for Medical Imaging, University College London, London, United Kingdom |
||
Magnetic Resonance Imaging is increasingly used for assessment of placental complications1. Here we apply MRI to examine Placenta Accreta Spectrum (PAS) a condition where the placenta is abnormally adherent or invasive. Super-Resolution Reconstruction (SRR) provides a high-resolution 3D reconstruction from 2D MRI slices to improve the visibility of structures for future clinical use2. Image Quality metrics allows quantitative evaluation of the SRR images to compare with the original 2D images. These metrics are found to be statistical significant, providing an objective assessment of the SRR images. |
| 973 | Computer 93
|
NOE imaging in tumors using asymmetric analysis of 2pi-CEST signals |
| Jing Cui1,2 and Zhongliang Zu1,2 | ||
1Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, 2Department of Radiology and Radiological Sciences, Vanderbilt University Medical Center, Nashville, TN, United States |
||
Nuclear overhauser enhancement (NOE) mediated chemical exchange saturation transfer (CEST) effect at -3.5 ppm has shown clinical interests in diagnosing tumors. Asymmetric analysis is a fast and simple method to measure and quantify the NOE effect, but has contamination from the downfield amide proton transfer (APT) effect at 3.5 ppm. In this abstract, we provide a new NOE quantification method using asymmetric analysis of CEST signals acquired with 2pi saturation pulses (2pi-CEST) which is also fast but has reduced contamination from the APT effect. |
| 974 | Computer 94
|
Refined feature selection method for the depiction of pathology by texture analysis – Applicated on MR images of intervertebral discs fissures |
| Christian Waldenberg1,2, Stefanie Eriksson1,2, Hanna Hebelka1,2, Helena Brisby1,2, and Kerstin Lagerstrand1,2 | ||
1University of Gothenburg, Gothenburg, Sweden, 2Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden |
||
Texture analysis combined with attention mapping has the potential to identify the position of pathology normally not visible in MR images by exploiting inter-pixel relationships in magnetic resonance (MR) images. However, as pathology can influence adjacent tissue, the features used to identify the position of the pathology have to be selected with care. Here we present an easily implemented method that efficiently selects texture features that are only sensitive to the pathology and can improve the localization of pathology even when not visible in MR images. |
| 1032 | Computer 39
|
SLOW-editing for GABA/GABA+ MRSI at 7T |
| Guodong Weng1,2, Piotr Radojewski1,2, and Johannes Slotboom1,2 | ||
1Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging (SCAN), University of Bern, Bern, Switzerland, 2Translational Imaging Center, sitem-insel AG, Bern, Switzerland |
||
Purpose: whole brain in vivo detection of GABA/GABA+ using SLOW-editing at 7T. Methods: An EPSI-based B0/B1+ robust SLOW-editing was applied in four healthy subjects. Pure GABA was measured by two different macromolecular-nulling inversion recovery pulses, namely a broadband- and a narrowband-inversion pulse. Results: the editing GABA/GABA+ signal at 3.01 ppm are presented. The signal intensity of GABA/GABA+ ratio is 40 – 60%. Conclusions: whole-brain in vivo GABA/GABA+-editing can be performed using SLOW-EPSI in around 10 minutes measurement time and is therefore clinically applicable. |
| 1033 | Computer 40
|
Simultaneous J-difference editing of GABA and GSH with extended echo-times using ExTE-HERMES |
| Peter Truong1 and Jamie Near1,2 | ||
1Physical Sciences Research Platform, Hurvitz Brain Sciences Program, Sunnybrook Research Institute, Toronto, ON, Canada, 2Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada |
||
J-difference editing techniques, such as MEGA-PRESS, are often used to detect gamma-aminobutryic acid (GABA) and glutathione (GSH). However, two separate scans are needed to acquire both metabolites. The HERMES editing scheme does address this issue, allowing simultaneous acquisition of GABA and GSH. Even then, there are limitations as the optimal echo time (TE) can only be set for one metabolite at the expense of lower editing efficiency for the other. We propose the Extended TE (ExTE) HERMES editing scheme, which allows us to acquire edited spectra at TE=120ms, ensuring maximum GSH signal while still providing efficient GABA editing. |
| 1034 | Computer 41
|
Cerebral Blood Flow Quantification using Pseudo-continuous Arterial Spin Labeling at 7T MRI |
| Salem Alkhateeb1, Tae Kim2, Howard J Aizenstein2, and tamer S. Ibrahim3 | ||
1University of Pittsburgh, PITTSBURGH, PA, United States, 2University of Pittsburgh, Pittsburgh, PA, United States, 3University of Pittsburgh, pittsburgh, PA, United States |
||
The feasibility of implementing pCASL technique at 7T MRI as a routine protocol uncovers the difficulties and challenges that need to be overcome and taking into consideration the many variables that need to be addressed before running the sequence. We were able to successfully implement pCASL sequence with satisfactory results. CBF quantification for the 4 volunteers was within clinical normal ranges and CBF mapping agrees with published data. Future work will include the inclusion of neck coil module for improving labeling efficiency.[GU1] Where is figure 4 referenced [GU1] Please add the references |
| 1035 | Computer 42
|
Prospective correction of multi-shot diffusion imaging using motion compensation and dual-speed EPI |
| Kevin Moulin1,2, Tyler E Cork3,4, Thomas Troalen5, Daniel Ennis3,4, Pierre Croisille1,2, and Magalie Viallon1,2 | ||
1CREATIS, Lyon, France, 2Department of Radiology, University Hospital Saint-Etienne, Saint Etienne, France, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4Division of Radiology, Veterans Administration Health Care System, Palo Alto, CA, United States, 5Siemens Healthcare SAS, Saint-Denis, France |
||
Multi-shot (MS) reduces image distortion in EPI caused by B0 field inhomogeneity and enables higher resolution diffusion imaging. However, interaction between physiologic motion and diffusion encoding gradients leads to strong image aliasing artifacts in MS-EPI due to phase discrepancies. We propose prospective strategies to reduce shot-to-shot phase variation for interleaved phase-segmented diffusion MS-EPI. First and second order motion compensated diffusion gradient waveforms and dual-speed EPI with oversampled k-space center were evaluated as a shot-to-shot phase mitigation strategy. Validations were performed in silico, in vitro, and in vivo in the brain, liver, and heart. |
| 1036 | Computer 43
|
Highly Asymmetric Acquisitions for Diffusion MRI Tractography |
| Manoj Shrestha1, Pavel Hok1,2, Ulrike Nöth1, and Ralf Deichmann1 | ||
1Brain Imaging Center (BIC), Goethe University Frankfurt, Frankfurt am Main, Germany, 2Department of Neurology, Palacky University Olomouc and University Hospital Olomouc, Olomouc, Czech Republic |
||
Frequently, diffusion-weighting modules such as schemes based on the Stejskal-Tanner concept or the STEAM are combined with EPI readout for fast k-space sampling. In contrast to standard sequences, highly asymmetric spin-echo and highly asymmetric STEAM sequences (HASE, HASTEAM) offer shorter TE, maintaining constant diffusion preparation duration, independent of the matrix-size. In this study, HASE and HASTEAM sequences were implemented for neuronal fibre tracking, and results were compared with standard sequences. Both asymmetric sequences HASE and HASTEAM allow for more accurate estimations of the subsidiary fibre orientations, which may considerably improve the quality and quantity of the detected fibre tracts. |
| 1037 | Computer 44
|
Investigating Spiral Arterial Spin Labeling with Pulseq and Field Monitoring at 7T |
| Ruoyun Emily Ma1, Dimo Ivanov1, Renzo Huber1, Denizhan Kurban1, and Benedikt A Poser1 | ||
1Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, Netherlands |
||
We developed several ASL sequences at 7T with a FAIR-QUIPSS II labeling scheme and various spiral readout strategies using Pulseq. Iterative algebraic image reconstruction was performed with CG-SENSE, using the field evolution data measured with external NMR probes. Robust performance in detecting brain’s perfusion signal was observed in 2D single- and multi-band spiral acquisitions especially at relatively high spatial resolution, without the requirement for a longer scan time. 3D spiral acquisition showed reduced contrast level in perfusion maps and requires further investigation and optimization. |
| 1038 | Computer 45
|
3D EPI sequence for indirect detection of hyperpolarized [2-13C]lactate with J-coupling artifact correction |
| Tyler Blazey1 and Cornelius von Morze1 | ||
1Washington University in St. Louis, St. Louis, MO, United States |
||
Imaging of hyperpolarized (HP) 13C magnetization is typically done using direct detection of the 13C signal. However, indirect 1H-based detection via J-coupled protons has the advantage of reducing gradient requirements and enhancing sensitivity, as well as practical advantages. We have developed an INEPT-based EPI sequence, which we tested by acquiring 3D images of a 13C formate phantom after transfer of thermal polarization from 1H to 13C. We also tested a novel deconvolution method to correct ghosting artifacts introduced in EPI images by J-coupling. These methods will be translated into in vivo HP imaging experiments in the next phase of research. |
| 1039 | Computer 46
|
Robust and Motion-Insensitive Approach to Diffusion-Weighted Half-Fourier Acquisition Single-Shot Turbo Spin-Echo Imaging |
| Aidin Arbabi1, Vitaliy Khlebnikov1, David. G. Norris1, and Jose P. Marques1 | ||
1Donders Institute, Radboud University, Nijmegen, Netherlands |
||
Half-Fourier acquisition single-shot turbo spin-echo diffusion-weighted (DW-HASTE) is a clinically established magnetic resonance imaging tool for the detection of small lesions, particularly cholesteatoma. However, in the standard approach, half of the available signal is sacrificed through displacing one echo parity out of the acquisition window to fulfil the Carr-Purcell-Meiboom-Gill condition. We present a selective parity approach to tackle this problem. The proposed method features a near full sensitivity, a low specific absorption rate for long echo trains, and about two-fold increase in signal-to-noise ratio, compared to the manufacturer's method under the same conditions. |
| 1040 | Computer 47
|
Pulmonary imaging Using 3D Dual-Echo FID Ultra-short Echo Time MRI with Rosette k-space Pattern: Introduction and Feasibility. |
| Nathan Ooms1, Xin Shen2, Ali Caglar Ozen3, Serhat Ilbey3, Mark Chiew4, and Uzay Emir1 | ||
1School of Health Sciences, Purdue University, West Lafayette, IN, United States, 2Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States, 3Radiology, Medical Center- University of Freiburg, Freiburg, Germany, 4Welcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom |
||
Pulmonary imaging traditionally uses techniques utilizing ionizing radiation such as x-rays and CT. Pediatrics, pregnant patients, and patients who will be exposed multiple times over long periods of time would benefit from a non-ionizing modality. MRI does not use ionizing radiation but suffers from low signal from the lungs and breathing motion artifact. We are proposing a novel MRI technique to combat these issues: a 3-dimensional Dual-Echo FID Ultra-short Echo Time (3D DE UTE) sequence utilizing Rosette k-space acquisition. Preliminary results demonstrated better pulmonary artery segmental branches and pleural wall definition compared to other commercially available MRI sequences. |
| 1041 | Computer 48
|
Simultaneous ADC mapping and water-fat separation with B0 correction using a rosette acquisition |
| Kang Yan1,2, Huajun She2, and Yiping Du2 | ||
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, 2Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China |
||
In diffusion-weighted imaging (DWI) using echo-planar acquisition, fat signals are commonly suppressed by using chemical saturation or by spectral-spatial RF excitation. In this study, we present a technique using a rosette trajectory for simultaneous mapping of apparent diffusion coefficient (ADC) and water-fat separation with inherent B0 correction. The feasibility of using this technique is demonstrated in phantom, brain, and head-neck scans. |
| 1042 | Computer 49
|
SLOW-EPSI for the Whole Brain Detection of Phosphorylethanolamine (PE) at 7T |
| Guodong Weng1,2, Piotr Radojewski1,2, Philippe Schucht3, Roland Wiest1,2, and Johannes Slotboom1,2 | ||
1Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging (SCAN), INSEL Gruppe AG, University of Bern, Bern, Switzerland, 2Translational Imaging Center, sitem-insel AG, Bern, Switzerland, 3Department of Neurosurgery, Inselspital Bern and University Hospital, Bern, Switzerland |
||
Purpose: to detect phosphoethanolamine (PEtn or PE) in vivo for whole-brain MRSI, overcoming challenges of B1+-inhomogeneities and CSDE at 7T. Methods: An EPSI-based B1+ robust SLOW-editing was applied to three healthy subjects. TE = 90 ms, TR = 1500 ms, and TA = 9:04 mins. Results: the editing PE signal at 3.22 ppm was clearly present, and the B1+-homogenous result/map was obtained. Conclusions: whole-brain in vivo PE-editing can be performed in around 9 minutes measurement time and is therefore clinically applicable. |
| 1043 | Computer 50
|
A modular software “sequence building block” for interleaved MRS and MRI acquisition on 7T Terra |
| Jabrane Karkouri1, Christopher Wickens1, Daniel Atkinson1, and Christopher T. Rodgers1 | ||
1University of Cambridge, Cambridge, United Kingdom |
||
In this work, we present a software “sequence building block” that enables interleaving MRS acquisitions (X-nuclear or 1H) inside a base imaging sequence (typically 1H) on the Siemens 7T Terra platform. We envisage using this approach for dynamic B0 shim and frequency offset correction to mitigate breathing artefacts in 31P-MRSI of heart and liver. |
| 1044 | Computer 51
|
Dual-encoding of magnetization transfer and diffusion for the characterization of tract-specific myelination |
| Ilana R Leppert1, Christopher D Rowley1,2, Jennifer SW Campbell1, Mark Nelson1,3, Bruce G Pike4, and Christine L Tardif1,2,3 | ||
1McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, Montreal, QC, Canada, 2Department of Biomedical Engineering, McGill University, Montreal, QC, Canada, 3Department of Neurology and Neurosurgery, McGill University, Montreal, QC, Canada, 4Hotchkiss Brain Institute and Departments of Radiology and Clinical Neuroscience, University of Calgary, Calgary, Canada, Calgary, AB, Canada |
||
Most white matter voxels are comprised of a combination of microstructurally different tracts that kiss and cross. Conventional MRI contrasts that provide a single measurement per voxel fail to disentangle the microstructural properties of these tracts. Co-encoding diffusion with a complementary contrast offers additional insight into each tract’s individual contribution. Here we present a novel MT-prepared diffusion sequence, optimized for contrast in both MTR and orientation, acquired within a clinically acceptable scan time of under 7 min. The goal is to provide a data acquisition framework for the computation of tract-specific myelin indices and potentially more structurally specific connectomes. |
| 1045 | Computer 52
|
Susceptibility-induced B-field inhomogeneity of undulating axons and effects on the DWI signal |
| Sidsel Winther1,2, Henrik Lundell2, and Tim B. Dyrby1,2 | ||
1Department of Applied Mathematics and Computer Science, Technical University of Denmark, Kongens Lyngby, Denmark, 2Danish Research Centre for Magnetic Resonance (DRCMR), Copenhagen University Hospital Hvidovre, Hvidovre, Denmark |
||
Magnetic susceptibility induces morphology- and orientation-dependent perturbations of the $$$B_0$$$-field and hence the DWI signal. At the micro-structural scale of brain white matter, the main contribution to these perturbations comes from myelin. To understand how the perturbations are shaped by the intrinsically undulating morphology of axons, observed in 3D reconstructions of synchrotron imaging and electron microscopy, we have implemented a framework that enables us to systematically study how undulation affects the orientation-dependency of susceptibility-induced B-field inhomogeneity, and how this affects the DWI signal - both in the intra- and extra-axonal compartments. |
| 1046 | Computer 53
|
Outlier Rejection for Fetal Brain MRS |
| Jack Highton1, Maria Yanez Lopez1, Anthony Price1, Vanessa Kyriakopoulou1, Lisa Story1, Mary A. Rutherford1, Jo Hajnal1, and Enrico De Vita1 | ||
1Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom |
||
Magnetic resonance spectroscopy (MRS) can measure the concentration of a range of metabolites to understand pathophysiological and normal development in the fetus, as well as provide diagnostic and prognostic markers [1]. However, MRS applications in the fetus are technically very challenging due to the inherently low SNR (fetal size, distance from the coil, the use of 1.5T scans) and signal disruption caused by unrestricted fetal motion during data collection. Given the limited available SNR, outlier rejection to remove motion corrupted averages and improve spectral quality is particularly important. Spectral phase shifts are associated with motion, both during signal acquisition and between MRS repeats [2]. We propose a threshold-free method for optimal outlier rejection of fetal MRS, “Ranked residual-water Phase Shift”: the phase shift in the residual water after water suppression is used as a metric linked to potential motion in individual transients/repeats. |
| 1047 | Computer 54
|
Optimization of a breast EPI-DWI protocol with high b-values and anterior-posterior phase-encoding direction: initial results |
| Ana E Rodríguez-Soto1, Summer J Batasin1, Lauren K Fang1, Michael Carl2, Haydee Ojeda-Fournier1, Kathryn E Keenan3, and Rebecca A Rakow-Penner1,4 | ||
1Radiology, University of California, San Diego, La Jolla, CA, United States, 2GE Healthcare, University of California, San Diego, La Jolla, CA, United States, 3National Institute of Standards and Technology, Boulder, CO, United States, 4Bioengineering, University of California, San Diego, La Jolla, CA, United States |
||
Diffusion-weighted imaging (DWI) acquired with echo-planar imaging (EPI) is utilized in breast MRI protocols to evaluate cancer lesions. While clinical implementation is limited by the distortion artifacts which affect EPI, methods such as reverse polarity gradient (RPG) and data collection with reduced field-of-view (FOV) have been used to reduce distortion. Evaluating these methods in minimizing distortion for images acquired with parallel imaging (PI) is a promising approach to improve EPI-DWI breast imaging. Here, we evaluated the distortion magnitude of EPI-DWI data collected with and without PI on a breast phantom and tested the feasibility of this protocol in vivo. |
| 1048 | Computer 55
|
Evaluating Within-Subject Test-Retest Replicability of Free-Water Corrected and Uncorrected Multi-Shell Diffusion MRI Measures. |
| Virendra R Mishra1, Ofer Pasternak2, Karthik R Sreenivasan1, and Dietmar Cordes1 | ||
1Imaging, Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, NV, United States, 2Harvard Medical School, Boston, MA, United States |
||
Free-water (FW) estimation could be improved with higher spatial resolution diffusion MRI (dMRI) data acquired at multiple shells. However, the effect of multi-shell protocol features on the accuracy of estimating FW and FW-corrected fractional anisotropy (FA) across the brain structures is currently unknown. We evaluated test-retest reproducibility of FW and FW-corrected FA across 20 major white-matter tracts across four dMRI protocols, two protocols with ADNI-3 and two protocols with HCP sequences. Our analysis suggests higher spatial resolution HCP-style dMRI data acquisition with correction of FW-estimation could be reliably performed in routine clinical investigations. |
| 1049 | Computer 56
|
MRSI Processing and Simulation Toolbox in the FID Appliance (FID-A) |
| Brenden Kadota1, Bernhard Strasser2,3, Wendy Oakden1, and Jamie Near1,4 | ||
1Physical Studies Research Platform, Sunnybrook Research Institute, Toronto, ON, Canada, 2Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria, 3A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, United States, 4Medical Biophysics, University of Toronto, Toronto, ON, Canada |
||
Magnetic resonance spectroscopic imaging (MRSI) is a biomedical imaging tool that enables spatial mapping of metabolite signals in vivo. Dedicated software for MRSI data preprocessing, image reconstruction, visualization and simulation is vital to the development and routine application of MRSI but relatively few tools are available for this purpose. We present an open source toolbox in MATLAB for preprocessing, visualization and simulation of MRSI data. The processing toolbox includes functions for importing data, coil combination, B0 inhomogeneity correction, and non cartesian reconstruction, while the simulation toolbox can simulate MRSI acquisitions with arbitrary input k-space trajectories. |
| 1050 | Computer 57
|
Protocol harmonization using a generative adversarial network decreases morphometry variability |
| Veronica Ravano1,2,3, Jean-François Démonet2, Daniel Damian2, Reto Meuli2, Gian Franco Piredda1,2,3, Till Huelnhagen1,2,3, Bénédicte Maréchal1,2,3, Jean-Philippe Thiran2,3, Tobias Kober1,2,3, and Jonas Richiardi2 | ||
1Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, 2Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 3LTS5, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland |
||
In radiology, the deployment of automated clinical decision support tools to new institutions is often hindered by inter-site data variability. In MRI, data heterogeneity often arises from differences in acquisition protocols. To overcome this issue, we propose a post-hoc harmonization technique based on generative adversarial networks (GAN). Seventy-seven patients suffering from dementia were scanned with two distinct T1-weighted MP-RAGE protocols. We show that cross-protocol harmonization of brain images using a conditional GAN improves image similarity and reduces the variability of brain morphometry. |
| 1051 | Computer 58
|
Machine Enhanced Reconstruction Learning and Interpretation Networks (MERLIN) |
| Kerstin Hammernik1,2 and Thomas Küstner3 | ||
1Lab for AI in Medicine, Technical University of Munich, Munich, Germany, 2Department of Computing, Imperial College London, London, United Kingdom, 3Medical Image And Data Analysis (MIDAS.lab), Department of Interventional and Diagnostic Radiology, University Hospital of Tuebingen, Tuebingen, Germany |
||
Machine Learning (ML) methods have evolved tremendously during the last decade. A number of frameworks support the development of new ML methods. However, support for high-dimensional and complex-valued data processing is often limited. Therefore, we propose a Machine Enhanced Reconstruction Learning and Interpretation Networks (MERLIN) framework that seamlessly integrates with existing ML solutions such as Tensorflow/Keras and Pytorch and complements them by high-dimensional, complex-valued and MR-specific operators and layers. Furthermore, standard data processing pipelines in Python are provided in the context of MRI reconstruction. |
| 1052 | Computer 59
|
A data-driven method for automatic regularization selection in a hybrid DL-SENSE reconstruction |
| Zahra Hosseini1, Thorsten Feiweier2, John Conklin3, Stephan Kannengiesser2, Marcel Dominik Nickel2, Min Lang3, Azadeh Tabari3, Augusto Lio Concalves Filho3, Wei-Ching Lo4, Maria Gabriela Figueiro Longo3, Michael Lev3, Pamela Schaefer3, Otto Rapalino3, Susie Huang3, Stephen Cauley5, and Bryan Clifford4 | ||
1MR R&D Collaboration, Siemens Medical Solutions USA, Atlanta, GA, United States, 2Siemens Healthcare GmbH, Erlangen, Germany, 3Department of Radiology, Massachusetts General Hospital, Boston, MA, United States, 4MR R&D Collaboration, Siemens Medical Solutions USA, Boston, MA, United States, 5Department of Radiology, A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States |
||
The integration of deep learning priors into regularized CG-SENSE reconstructions enables high quality MR images to be generated from noisy, undersampled data. The regularization parameter in these methods can be tuned to control the level of denoising, allowing a network to generalize to novel SNR conditions without retraining. However, manual tuning of the regularization parameter can be time consuming. This work presents a data-driven method for automatic regularization selection using commonly acquired noise calibration data. Results indicate the method generalizes across clinically relevant imaging scenarios and provides diagnostically equivalent image quality to that obtained by manual parameter tuning. |
| 1053 | Computer 60
|
Iterative RAKI with complex-valued convolution for improved image reconstruction with limited training samples |
| Peter Dawood1,2, Martin Blaimer3, Felix Breuer3, Paul R. Burd4, István Homolya5,6, Peter M. Jakob1, and Johannes Oberberger2 | ||
1Department of Physics, University of Würzburg, Würzburg, Germany, 2Department of Internal Medicine I, University Hospital Würzburg, Würzburg, Germany, 3Magnetic Resonance and X-ray Imaging Department, Fraunhofer IIS, Fraunhofer Institute for Integrated Circuits IIS, Division Development Center X-Ray Technology, Würzburg, Germany, 4Institute for Theoretical Physics and Astrophysics, University of Würzburg, Würzburg, Germany, 5Brain Imaging Centre, Research Centre for Natural Sciences, Budapest, Hungary, 6Institute of Nuclear Techniques, Budapest University of Technology and Economics, Budapest, Hungary |
||
Recently, the Parallel Imaging method GRAPPA has been generalized by the deep-learning method RAKI, in which Convolutional Neural Networks are used for non-linear k-space interpolation. RAKI uses scan-specific training data, however, due to its increased parameter-space, its reconstruction quality may deteriorate given a limited training-data amount. We evaluate an approach that includes augmented training-data via an initial GRAPPA k-space reconstruction, and weights refinement by iterative training. Thereby, severe residual artefacts are suppressed in RAKI, while preserving its resilience against g-factor noise enhancement in GRAPPA for standard 2D imaging at medium accelerations, for strongly varying contrast between training- and interpolation-data, too. |
| 1054 | Computer 61
|
Data scarcity mitigation approaches in deep learning reconstruction of undersampled low field MR images |
| Tobias Senft1, Reina Ayde1, Najat Salameh1, and Mathieu Sarracanie1 | ||
1Department of Biomedical Engineering, University of Basel, Allschwil, Switzerland |
||
Low magnetic field (LF) MRI is gaining popularity as a flexible and cost-effective complement to conventional MRI. However, LF-MRI suffers from a low signal-to-noise ratio per unit time which calls for signal averaging and hence prolonged acquisition times, challenging the clinical value of LF MRI. In this study, we show that Deep Learning (DL) can reconstruct artifact-free heavily undersampled 2D LF MR images (34% sampling) with great success, both retrospectively and prospectively. Our results also highlight that a transfer learning approach combined with data augmentation improves the overall reconstruction performances, even when only small LF training datasets are available. |
| 1055 | Computer 62
|
Concept of a symmetry-guided single-layer neural network for image reconstruction of undersampled radial-MRI k-space data. |
| Sjoerd Ypma1, Ivo Maatman1, Matthan Caan2, Dimitris Karkalousos2, Marnix Maas1, and Tom Scheenen1 | ||
1Radboud UMC, Radiology and Nuclear Medicine, University of Nijmegen, Nijmegen, Netherlands, 2Amsterdam UMC, Biomedical Engineering and Physics, University of Amsterdam, Amsterdam, Netherlands |
||
Undersampled k-space data reconstruction results in aliasing artifacts. Compressed sensing theory enables image reconstruction by using a priori knowledge in the form of regularization. Increasingly, Machine Learning methods are used to learn the regularization from data itself, but these methods can result in unstable reconstructions. We propose a translation equivariant single-layer neural network for reconstruction of radially measured k-space data. By exploiting translation symmetry, it can learn from randomly simulated data while still being applicable to in-vivo measurements. We tested robustness to small perturbations and reliability of the reconstruction of unexpected objects. |
| 1056 | Computer 63
|
Accelerating MR Elastography using Deep Learning-Reconstruction of Undersampled Data |
| Robin Antony Birkeland Bugge1, Jon Andre Ottesen2, Elies Fuster3, Atle Bjørnerud2, and Kyrre Eeg Emblem1 | ||
1Department of Diagnostic Physics, Oslo University Hospital, Oslo, Norway, 2Department of Computational Radiology and Artificial Intelligence, Oslo University Hospital, Oslo, Norway, 3Biomedical Data Science Laboratory, Instituto Universitario de Tecnologías de la Información y Comunicaciones, Universitat Politècnica de València, Valencia, Spain |
||
Problem summary: Brain MR elastography is associated with extended acquisition times, which is alleviated by reduced coverage or resolution. The aim of this project is to utilize deep learning to accelerate MRE. Methods: We employ an MRE for fully sampled acquisition. Undersampled data is simulated by masking phase-encoding steps. A cascaded reconstruction network is used to reconstruct the phase image from undersampled k-space. Results: There are subtle differences between the reconstructed and fully sampled phase images. We observe a non-significant difference for stiffness values in our preliminary results. Conclusions: The method shows promise for accelerating MR elastography data. |
| 1057 | Computer 64
|
Brain MR image super resolution using simulated data to perform in real-world MRI |
| Aymen Ayaz1, Kirsten Lukassen1, Cristian Lorenz2, Juergen Weese2, and Marcel Breeuwer1,3 | ||
1Biomedical Engineering Department, Eindhoven University of Technology, Eindhoven, Netherlands, 2Philips Research Laboratories, Hamburg, Germany, 3MR R&D – Clinical Science, Philips Healthcare, Best, Netherlands |
||
We propose to simulate a large set of anatomically variable voxel-aligned and artifact-free brain MRI data at different resolutions to be used for training deep-learning based Super Resolution (SR) networks. To the best of our knowledge, no such efforts have been made in past regarding use of simulated data to train a SR network. We trained a SR network using such simulated data and tested the performance on real-world MRI data. The trained network could significantly sharpen low-resolution input MR images and clearly outperformed classic image interpolation methods. |
| 1058 | Computer 65
|
Complex graph convolutional neural networks compared to GRAPPA for reconstruction of undersampled non-Cartesian MRSI |
| Paul Weiser1, Stanislav Motyka2, Wolfgang Bogner2, and Georg Langs1 | ||
1Computational Imaging Research Lab - Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 2High Field MR Center - Department of Biomedical Imaging and Image‐Guided Therapy, Medical University of Vienna, Vienna, Austria |
||
In this work a Graph Neural Network for the reconstruction of undersampled k-space data is proposed. The results were evaluated and compared against a state-of-the-art method. Overall, the results suggest that this Deep Learning-based approach for reconstruction is promising. |
| 1059 | Computer 66
|
Signal prediction in echo dimension of multi-echo gradient echo using multi-layer seq2seq model |
| Jisu Yun1, Seul Lee1, Muyul Park1, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of |
||
The mGRE(multi-echo gradient echo) sequence has previously been used for MWF(Myelin water fraction) imaging. Such mGRE observes signal decay by obtaining multiple echo signals, but scan time increases as more echoes are obtained. To solve this trade-off, we developed a deep learning model based on a LSTM model that can reduce scan time by predicting the later echoes using only the early echoes. Looking at the in vivo results and various performance test, our network has lower RMSE and higher PSNR than NLLS(Non-linear least squares), a conventional fitting algorithm. |
| 1060 | Computer 67
|
Deep learning reconstruction enables accelerated acquisitions with consistent volumetric measurements |
| R. Marc Lebel1, Suryanarayanan Kaushik2, Trevor Kolupar2, Nate White3, Weidong Luo3, and Suchandrima Banerjee4 | ||
1GE Healthcare, Calgary, AB, Canada, 2GE Healthcare, Waukesha, WI, United States, 3Cortechs, San Diego, CA, United States, 4GE Healthcare, Menlo Park, CA, United States |
||
This work evaluates brain volumetry measurements obtained from T1w images with fast (PI) and rapid (PI+CS) protocols (< 3 mins) and reconstructed with AIR Recon DL 3D, a deep learning-based reconstruction to reduce noise and ringing. Images were segmented using FDA-cleared software NeuroQuantÒ for quantitative volumetric analysis. Segmentation was successful in all cases and key brain volumes are unchanged between fast and rapid protocols, and further between conventional and DL reconstructions. This work demonstrates that highly accelerated acquisitions and advanced reconstruction methods are suitable for segmentation and volumetric studies and can improve repeatability of measurements. |
| 1061 | Computer 68
|
Single-Shot Adaptation using Score-Based Models for MRI Reconstruction |
| Marius Arvinte1, Ajil Jalal1, Giannis Daras2, Eric Price2, Alex Dimakis1, and Jonathan I Tamir1,3,4 | ||
1Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, United States, 2Computer Science, The University of Texas at Austin, Austin, TX, United States, 3Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin, Austin, TX, United States, 4Diagnostic Medicine, Dell Medical School, The University of Texas at Austin, Austin, TX, United States |
||
This work deals with the problem of few-shot adaptation in data-driven MRI reconstruction, where models must efficiently adapt to new distributions. We introduce score-based models for MRI reconstruction and an algorithm for adjusting inference parameters (step size, noise level and stopping point), investigate the impact of these parameters on reconstruction performance, and demonstrate average gains of at least 2 dB in PSNR across a range of acceleration values, all while using a pretrained model that was trained for brain MRI and fine-tuned using only a single fully-sampled 2D knee scan from the fastMRI database. |
| 1142 | Computer 47
|
Impact of the B0 and B1+-adjustments on the in vivo metabolite quantification accuracy of simultaneous 2voxel 1H brain MRS at 7T |
| Layla Tabea Riemann1, Christoph Stefan Aigner1, Ralf Mekle2, Sebastian Schmitter1,3,4, Bernd Ittermann1, and Ariane Fillmer1 | ||
1Physikalisch-Technische Bundesanstalt (PTB), Berlin, Germany, 2Center for Stroke Research Berlin, Charité Universitätsmedizin, Berlin, Germany, 3Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States, 4Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany |
||
For simultaneous multi-voxel spectroscopy (sMVS), it is necessary to optimize the B0 shimming and the B1+ adjustment simultaneously for two voxels. The impact of these adjustments on the smallest possible distance for simultaneous two-voxel MRS acquisitions is determined by Bloch simulations, and the influence on the spectral quality is assessed for three different brain regions. To this end, the previously introduced 2 spin-echo full-intensity acquired localization (2SPECIAL) sequence and the voxel-GeneRalized Autocalibrating Partial Parallel Acquisition (vGRAPPA) decomposition algorithm are utilized to simultaneously acquire and retrospectively decompose in vivo brain 1H-MRS data from two voxels at short echo times at 7T. |
| 1143 | Computer 48
|
Quantitative measurements of GABA in mouse brain using MR Spectroscopy at 7T |
| Mohamed Tachrount1, Jason Lerch1,2, and Stuart Clare1 | ||
1Wellcome Centre for Integrative Neuroimaging (FMRIB), Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, 2Mouse Imaging Centre (MICe), University of Toronto, Toronto, ON, Canada |
||
The goals of this study were to implement a MEGA-sLASER sequence with prospective field drift correction on a preclinical scanner, develop the processing tools and validate them on a group of wild type mice. |
| 1144 | Computer 49
|
Comparison of pCASL, DSC and FAIR for Cerebral Perfusion in Rats @9.4T |
| Lisa Y Hung1, Karl-Heinz Herrmann1, Renat Sibgatulin1, Lydiane Hirschler2,3, Jan Warnking3, Emmanuel L Barbier3, Otto W Witte4, Knut Holthoff4, Jürgen R Reichenbach1, and Alexander Joerk4 | ||
1Medical Physics Group, Institute of Diagnostic Radiology, Jena University Hospital, Jena, Germany, 2Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, 3Grenoble Institut Neurosciences, Univ. Grenoble Alpes, Inserm, U1216, Grenoble, France, 4Hans-Berger Department of Neurology, Jena University Hospital, Jena, Germany |
||
Attaining reproducible, time-efficient, high spatial resolution and quantitative MRI perfusion images in small animal MRI @9.4T is still challenging. To take a step toward this goal, perfusion data from arterial spin labeling (ASL), Flow-sensitive Alternating Inversion Recovery (FAIR), dynamic susceptibility contrast (DSC), and pseudo-continuous arterial spin labeling (pCASL) sequences are compared by assessing cerebral perfusion in 5 rats, each scanned 3 times. Differences in (cerebral blood flow) CBF in major brain regions were analyzed, and the repeatability of perfusion MRI was assessed. |
| 1145 | Computer 50
|
The impact of optimal RF coil-combination on whole-brain sub-millimetre resolution perfusion imaging at 7T |
| Sriranga Kashyap1,2, Dimo Ivanov2, Roy A. M. Haast3, Francisco J. Fritz4, Robbert L. Harms2, Benedikt A. Poser2, and Kamil Uludag5,6 | ||
1Techna Institute, University Health Network, Toronto, ON, Canada, 2Department of Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands, 3Aix-Marseille Universite, CNRS, CRMBM, Marseille, France, 4Institute for Systems Neuroscience, Universitätsklinikum Hamburg-Eppendorf, Hamburg, Germany, 5Techna Institute & Koerner Scientist in MR Imaging, University Health Network, Toronto, ON, Canada, 6Center for Neuroscience Imaging Research, Institute for Basic Science & Department of Biomedical Engineering, Sungkyunkwan University, Suwon, Korea, Republic of |
||
In this work, we use 3D-EPI ASL to acquire perfusion maps of the human brain at an unprecedented spatial resolution of 0.7 mm isotropic at 7T. This multi-session, single-subject ASL dataset offers a unique opportunity to investigate the cortical distribution of baseline perfusion across and within brain areas, as well as studying the physiological basis for the interpretation of laminar and columnar fMRI. In this abstract, we use an open-source, memory-efficient, CPU/GPU accelerated coil-combine Python toolbox to probe the impact of using covariance-weighted sum-of-squares (CovSoS) and tSNR optimised RF coil-combination (STARC) on high-resolution perfusion imaging at 7T. |
| 1146 | Computer 51
|
Multi-Shot Liver Diffusion MRI Using Variable Auto-Calibrating (vARC) Sampling Across Averages |
| Ashok Kumar P Reddy1, Harsh Kumar Agarwal1, Rajdeep Das1, Rajagopalan Sundaresan1, Shaik Ahmed1, Sajith Rajamani1, Bhairav Mehta1, Gaohong Wu1, M Ramasubba Reddy2, and Ramesh Venkatesan1 | ||
1GE Healthcare, Bangalore, India, 2Indian Institute of Technology Madras, Chennai, India |
||
Obesity is a key biomarker of liver pathology. Diffusion MRI in liver is key protocol in the absence on intravenous contrast agent and also provides additional insights on diffuse liver diseases and lesions However, its usage is limited by the inability to use multi-channel surface coil for obese liver patients in non-wide-bore MRI scanners. A variable k-space sampling scheme and auto-calibrating image reconstruction technique, vARC, is proposed in this abstract to reduce the amount of distortion and improve the image quality of Liver DWI acquired with single channel volume coil. |
| 1147 | Computer 52
|
Intra-Subject Variability of Skeletal Muscle Glycogen Using 13C/1H MRS at 3T Using a Novel Standardization Method |
| Rajakumar Nagarajan1, Gwenael Layec2, and Jane A Kent2 | ||
1Human Magnetic Resonance Center, Institute for Applied Life Sciences, University of Massachusetts Amherst, Amherst, MA, United States, 2Kinesiology, University of Massachusetts Amherst, Amherst, MA, United States |
||
Given its central role in substrate metabolism, the ability to quantify muscle glycogen (Gly) continuously and noninvasively using natural abundance 13C MR spectroscopy presents an important opportunity for understanding human metabolism. This study evaluated the reproducibility of Gly measured by 13C MRS in the quadriceps muscle using three methods for signal standardization at 3T: Gly/total creatine (TCr) from 13C, Gly/TCr using 13C and 1H, and Gly/water. The novel results from this study suggest that Gly/TCr measurements using the creatine peak from 1H MRS is highly reproducible and may be an advantageous approach to measuring muscle glycogen in the future. |
| 1148 | Computer 53
|
Diffusion-weighted MEGA-PRESS spectroscopy |
| Emile Schjeldsøe Berg1 and John Georg Seland1 | ||
1Department of Chemistry, University of Bergen, Bergen, Norway |
||
The editing capabilities of the MEGA-PRESS sequence are combined with diffusion-weighing to enable both quantification and diffusion measurements of less abundant brain metabolites. The technique, labelled DW-MEGA-PRESS, was tested for combined editing and diffusion-weighting of GABA. Obtained data give reliable results for GABA concentrations and corresponding diffusion coefficients from in vitro experiments. When applied to in vivo acquisitions, reliable diffusion coefficients for the dominating metabolites were obtained. GABA could be reliably quantified, but the signal-to-noise ratio was not sufficient for a reliable determination of its diffusion coefficient. This could be improved by further optimization of scanning and post-processing protocols. |
| 1149 | Computer 54
|
Frequency sweep 31P MR spectroscopic imaging at 7T |
| Songi Lim1,2, Mark Stephan Widmaier1,2, and Lijing Xin1,3 | ||
1CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 2Laboratory for Functional and Metabolic Imaging, EPFL, Lausanne, Switzerland, 3Animal Imaging and Technology, EPFL, Lausanne, Switzerland |
||
Phase-cycled spectroscopic imaging (PCSI) method was implemented and validated for 31P PCSI imaging at 7T. The PCSI method uses a balanced steady-state free precession sequence with an ultra-low flip angle (<1°) to achieve sharp passband with 2.52-ms of TR, which enable to accelerate the acquisition. With prior knowledge of 31P spectra, it is feasible to acquire major 31P peaks by changing the frequency offset and non-uniform phase sweeping instead of acquiring full spectra uniformly. To investigate feasibility of the method, a multi-compartment KH2PO4 phantom with in vivo equivalent concentrations was prepared and 31P PCSI was compared with conventional FID-CSI. |
| 1150 | Computer 55
|
A spectrally interleaved magnetic resonance spectroscopic imaging sequence incorporating semi-adiabatic pulses at ultrahigh field |
| Gaurav Verma1, Seena Dehkharghani2, Leeor Alon3, and Priti Balchandani1 | ||
1Biomedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2Radiology and Neurology, New York University, New York, NY, United States, 3Radiology, New York University, New York, NY, United States |
||
A spectrally-interleaved semi-adiabatic magnetic resonance spectroscopic imaging sequence was developed to simultaneously acquire water and metabolite resonances within the same repetition time. A water interleave enables eddy current correction, absolute quantification through water reference and thermometry using empirically-derived formulae based on chemical shifts of temperature-sensitive water and temperature-insensitive metabolites. The sequence successfully acquired both 1-minute single voxel and 4-minute multi-voxel acquisitions in phantom and in vivo, producing temperature estimates of 21.2 °C, and 35.7 °C, respectively. LCModel fitting of metabolites provided reliable fitting of multiple metabolite peaks including separation of glutamate and glutamine. |
| 1151 | Computer 56
|
Comparison of in vivo hepatic metabolic quantitation by localized 31P MRS at 7T |
| Lorenz Pfleger1,2, Wolfgang Bogner2, Albrecht Ingo Schmid3, Thomas Scherer1, Peter Wolf1, Michael Krebs1, and Martin Krššák1,4 | ||
1Division of Endocrinology and Metabolism, Department of Medicine III, Medical University of Vienna, Vienna, Austria, 2High Field MR Centre, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 3High Field MR Centre, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria, 4Christian Doppler Laboratory for Clinical Molecular MR Imaging, Vienna, Austria |
||
This study focuses on the comparison of different in vivo data-acquisitions for quantifying hepatic 31P metabolite concentrations using a phantom replacement method. We have compared a 1D-slab localized DRESS, a single-voxel-spectroscopy (SVS) ISIS as well as a 3D-CSI 31P MRS acquisition at 7T. Each method has its own advantages and disadvantages and needs to be chosen according to the focus and design of the study. |
| 1152 | Computer 57
|
Super-resolution mean diffusivity spectroscopic MRI in the human brain |
| Alexandru V Avram1,2, Magdoom Kulam1, Joelle E Sarlls3, Raisa Freidlin4, and Peter J Basser1 | ||
1Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, United States, 2Center for Neuroscience and Regenerative Medicine, The Henry Jackson Foundation, Bethesda, MD, United States, 3National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, United States, 4Center for Information Technology, National Institutes of Health, Bethesda, MD, United States |
||
We describe a comprehensive pipeline for super-resolution reconstruction of clinical diffusion-weighted MRIs acquired with isotropic diffusion encoding (IDE), or spherical tensor encoding. The pipeline integrates blip-up/down EPI distortion correction and slice-to-volume registration (SVR). Multiple low-resolution IDE-MRIs with different slice orientations relative to the brain are processed to reconstruct high-resolution IDE-MRIs. From high-resolution IDE-MRIs with a wide range of b-values we estimate spectra of subvoxel MD values to describe the distribution of water mobilities in microscopic brain tissue microenvironments. Integrating SVR-reconstruction with IDE is an important step in the clinical translation of MD spectroscopic MRI for fetal MRI applications. |
| 1153 | Computer 58
|
Simultaneous Measurement of Glu and GABA at 7T |
| Tal Finkelman1, Edna Furman-Haran2, Rony Paz3, and Assaf Tal1 | ||
1Chemical and Biological Physics, Weizmann, Rehovot, Israel, 2Life Sciences Core Facilities, Weizmann, Rehovot, Israel, 3Brain Sciences, Weizmann, Rehovot, Israel |
||
Glutamate (Glu) and γ‑aminobutyric-acid (GABA) are central neurotransmitters participating in many cognitive processes, and their combined measurement is of great relevance. Proton Magnetic Resonance Spectroscopy (1H-MRS) allows the no–invasive, in-vivo, measurement of GABA and Glu. At magnetic fields below 7T, GABA is detected by spectral editing strategies, while Glu is detected by non-editing strategies. We compared both edited and non‑edited approaches at 7T for simultaneously quantifying Glu and GABA from 21 volunteers, and found that non-edited sequences at TE=80ms provide better combined-reproducibility than either edited sequences at the same TE, or non-edited sequence at a shorter TE (42ms). |
| 1154 | Computer 59
|
Postprocessing Reduces Pulsation Artifacts and Increases Visibility of Liver Lesions in Flow-compensated Diffusion-weighted Imaging |
| Tobit Führes1, Marc Saake 1, Hannes Seuss1,2, Astrid Müller1, Sebastian Bickelhaupt1, Alto Stemmer3, Thomas Benkert3, Michael Uder1, Bernhard Hensel4, and Frederik Bernd Laun1 | ||
1University Hospital Erlangen, Erlangen, Germany, 2Klinikum Forchheim, Forchheim, Germany, 3Siemens Healthcare GmbH, Erlangen, Germany, 4Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany |
||
Diffusion-weighted imaging of the liver is prone to the cardiac pulsation artifact, which can lead to reduced lesion visibility. We addressed this problem with a two-fold approach. First, flow-compensated diffusion weightings were used, which are known to reduce this artifact. Using a dataset of 40 patients suffering from focal liver lesions, we addressed the remaining signal voids with different postprocessing techniques, namely weighted averaging, the p-mean approach, and an outlier exclusion algorithm. The algorithms substantially increased the lesion visibility and further reduced the pulsation artifact. An evaluation of CNR and calculation time showed that weighted averaging was suited best. |
| 1155 | Computer 60
|
Quality Assurance and Quality Control for HRMAS MRS Studies of Human Blood |
| Leo Cheng1, Matteo Sanchez-Dahl1, Isabella Muti1, JianXiang Weng1, and Anya Zhong1 | ||
1Massachusetts General Hospital, Charlestown, MA, United States |
||
The lack of a standard universal protocol for sample preparation in MRS experiments results in instrumental and biological variability across reported studies. The purpose of this experiment is to evaluate a quality assurance protocol for performing HRMAS 1H MRS experiments with human biofluid samples, such as blood serum. The research objective is to determine if the spectral results measured from pooled samples can be used to assess results from individual samples as controls. These findings will help better develop efficient standard protocols for researchers and clinicians to use in the future. |
| 1156 | Computer 61
|
Predicting isocitrate dehydrogenase mutation status using contrastive learning and graph neural networks |
| Yiran Wei1, Chao Li1, and Stephen John Price1 | ||
1Department of Clinical Neurosciences, University of Cambridge, Cambridge, United Kingdom |
||
The isocitrate dehydrogenase (IDH) gene mutation status is a prognostic biomarker for gliomas. As an alternative approach to the invasive gold standard of IDH mutation detection, radiogenomics approaches using features of MRI showed promising results in the same task. Here, we proposed an approach to predict the IDH status based on the image and morphology features extracted using contrastive learning. We then constructed a large patient graph based on the extracted features, which could predict the IDH mutation using graph neural networks. The results showed the proposed method outperforms the classifiers which leverage either image or morphology features only. |
| 1157 | Computer 62
|
Coil to Coil: Self-supervised denoising using phased-array coil images |
| Juhyung Park1, Dongwon Park2, Hyeong-Geol Shin1, Eun-Jung Choi1, Dongmyung Shin1, Se Yong Chun1, and Jongho Lee1 | ||
1Department of Electrical and Computer Engineering, Seoul National University, Seoul, Korea, Republic of, 2Ulsan National Institute of Science and Technology, Ulsan, Korea, Republic of |
||
A self-supervised learning framework, Coil to Coil (C2C), is proposed. This method generates two noise-corrupted images from single phased-array coil data to train a denoising network and, therefore, requires no clean image nor acquisition of a pair of noisy images. The two images are processed to have the same signals and independent noises, satisfying conditions for the noise to noise algorithm, which requires paired noise-corrupted images. C2C shows the best performance among popular self-supervised denoising methods in both real and synthetic noised images, revealing little or no structure in the noise map. |
| 1158 | Computer 63
|
Deep Learning-Based Receiver Coil Sensitivity Map Estimation for SENSE Reconstruction using Transfer Learning |
| Madiha Arshad1, Mahmood Qureshi1, Omair Inam1, and Hammad Omer1 | ||
1Medical Image Processing Research Group (MIPRG), Dept. of Elect. & Comp. Engineering, COMSATS University, Islamabad, Pakistan |
||
Many Parallel MRI algorithms (e.g. Sensitivity Encoding (SENSE)) require knowledge of the receiver coil sensitivity maps. Magnetic field strength is an important factor in defining the sensitivity maps of the receiver coils in MRI. This paper presents a method to estimate the receiver coil sensitivity maps of a higher magnetic field strength scanner utilizing a deep learning network (denoted as ResU-Net-34), initially trained on the receiver coil sensitivity maps of a lower field strength scanner using transfer learning. SENSE reconstruction results show a successful domain transfer between the receiver coil sensitivities of different magnetic field strengths with the proposed method. |
| 1159 | Computer 64
|
GRASPNET: Spatiotemporal deep learning reconstruction of golden-angle radial data for free-breathing dynamic contrast-enhanced MRI |
| Ramin Jafari1, Richard Kinh Gian Do1, Maggie Fung2, Ersin Bayram2, and Ricardo Otazo1 | ||
1Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2GE Healthcare, New York, NY, United States |
||
GRASP is a valuable tool to perform free-breathing dynamic contrast-enhanced (DCE) MRI with high spatial and temporal resolution. However, the 4D reconstruction algorithm is iterative and relatively long for clinical studies. In this work, we present a spatiotemporal deep learning approach to significantly reduce the reconstruction time without affecting image quality. |
| 1160 | Computer 65
|
Denoising Diffusion MRI with Self-supervised Learning on Coresets via Matrix Sketching |
| Shreyas Fadnavis1, Agniva Chowdhury2, Petros Drineas2, and Eleftherios Garyfallidis1 | ||
1Indiana University Bloomington, Bloomington, IN, United States, 2Purdue University, West Lafayette, IN, United States |
||
Diffusion MRI typically has a low SNR on account of the noise from a variety of sources corrupting the data. The state-of-the-art denoiser Patch2Self proposed a self-supervised learning technique that uses patches from all the voxels to learn the denoising function which in practice can be resource-intensive. We, therefore, propose Patch2Self2 which performs self-supervised denoising using coresets constructed via matrix sketching, resulting in significant speedups and reduced memory usage. Our results showed that sampling-based sketching via leverage scores gave the best performance. Remarkably, leverage scores can be directly used as a statistic for interpreting influential regions hampering the denoising performance. |
| 1161 | Computer 66
|
Super Resolution of MR Images with Deep Learning Based k-space Interpolation |
| Madiha Arshad1, Mahmood Qureshi1, Omair Inam1, and Hammad Omer1 | ||
1Medical Image Processing Research Group (MIPRG), Dept. of Elect. & Comp. Engineering, COMSATS University, Islamabad, Pakistan |
||
Super resolution of MR images can be used to speed up MRI scan time. However, super resolution is a highly ill-posed problem as the low-resolution images lack high frequency spatial information. In this paper, we propose a hybrid dual domain cascaded U-Net to restore the high-resolution images. Firstly, the U-Net operating in k-space domain is used to interpolate the missing k-space data points and then the U-Net operating in image domain provides a refined high-resolution solution image. Experimental results show a successful reconstruction of high resolution images by using only central 6.25% and 25% k-space data. |
| 1162 | Computer 67
|
Brain MRI acceleration with deep modular networks (BRACELET) |
| Anthony Mekhanik1, Robert Young2, and Ricardo Otazo1,2 | ||
1Dept. of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2Dept. of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States |
||
Brain cancer screening utilizing MRI suffers from a bottleneck by requiring 3D acquisitions before and after contrast injection. We designed a novel two-stage deep learning reconstruction pipeline to accelerate 3D brain MRI over conventional 2-fold parallel imaging accelerations used in clinical practice. Using modular deep neural networks, we removed the dependence on one all-encompassing network. We successfully dealiased brain images at higher accelerations and with structural fidelity in lesions superior to conventional clinical imaging. Our method was validated on pathologies unseen during training using qualitative evaluation from an expert neuroradiologist and achieved comparable scores to conventional 2-fold clinical accelerations. |
| 1163 | Computer 68
|
Applying advanced denoisers to enhance highly undersampled MRI reconstruction under plug-and-play ADMM framework |
| Kang Yan1, Zhixing Wang1, Quan Dou1, Sheng Chen1, and Craig H Meyer1,2 | ||
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, 2Radiology & Medical Imaging, University of Virginia, Charlottesville, VA, United States |
||
Accelerating MRI acquisition is always in high demand, since long scan time can increase the potential risk of image degradation caused by patient motion. Generally, MRI reconstruction at higher undersampling rates requires regularization terms, such as wavelet transformation and total variation transformation. This work investigates employing the plug-and-play (PnP) ADMM framework to reconstruct highly undersampled MRI k-space data with three different denoiser algorithms: block matching and 3D filtering (BM3D), weighted nuclear norm minimization (WNNM) and residual learning of deep CNN (DnCNN). The results show that these three PnP-based methods outperform current regularization methods. |
| 1164 | Computer 69
|
DL2: Robustness of Deep Learning Processing on Deep Learning reconstructed Lumbar Spine MRI |
| Madeline Hess1, Emma Bahroos1, Misung Han1, Kenneth Gao1, Valentina Pedoia1, and Sharmila Majumdar1 | ||
1University of California, San Francisco, San Francisco, CA, United States |
||
We demonstrate that fast acquisition images reconstructed robust to segmentation with convolutional neural networks, which are known to generate chaotic errors in response to minimal image perturbations. This finding indicates that low-NEX images with DL-based noise reduction are not only robust to human readers, but machine readers as well. |
| 1165 | Computer 70
|
Intrinsic reproducibility issues in deep learning-based MR reconstruction |
| Chungseok Oh1, Woojin Jung2, and Jongho Lee1 | ||
1Department of Electrical and Computer Engineering, Seoul National University, Seoul, Korea, Republic of, 2AIRS Medical Inc., Seoul, Korea, Republic of |
||
Deep learning requires a large number of parameter settings, which can be prone to reproducibility issues, undermining the reliability and validity of the outcomes. In this study, we list the common sources that may induce the reproducibility issues in deep learning-based MR reconstruction. The effect size of each source on the network performance was investigated. From the results of this study, we recommend to share a trained network. |
| 1166 | Computer 71
|
Sampling Pattern Optimization for Joint Acceleration of Multi-contrast MRI using Deep Learning |
| Sunghun Seo1, Huan Minh Luu1, Seung Hong Choi2, and Sung-Hong Park1 | ||
1Bio and Brain Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, 2Radiology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea, Republic of |
||
Usage of multiple-acquisition MRI is one field of study that proved its effectiveness and practicality since routine MR scan protocol typically acquires multiple information for the same anatomical structures. In this study, we propose simultaneous optimization of sampling pattern and reconstruction for joint acceleration of multi-contrast MRI. The simultaneous optimization of sampling pattern and reconstruction provided superior performance over single contrast imaging and over single sampling pattern for multi-contrast MRI. The proposed technique can be adopted in routine clinical scan without forcing extra scans during acquisition. |
| 1244 | Computer 40
|
Evaluation of quantitative T1 and PD mapping at 7T from an MP2RAGE Sequence optimised to obtain UNI and FLAWS contrast images in a single scan |
| Ayse Sila Dokumaci1,2, Katy Vecchiato2,3,4, Raphael Tomi-Tricot1,2,5, Pip Bridgen1,2, Michael Eyre1,2, Tobias C. Wood6, Chiara Casella2,4, Jan Sedlacik2,7, Tom Wilkinson1,2, Sharon Giles1,2, Joseph V. Hajnal1,2, Shaihan Malik1,2, Jonathan O’Muircheartaigh2,3,4,8, and David W. Carmichael1,2 | ||
1Biomedical Engineering Department, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2London Collaborative Ultra high field System (LoCUS), London, United Kingdom, 3Department of Forensic and Neurodevelopmental Sciences, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, United Kingdom, 4Centre for the Developing Brain, School of Biomedical Engineering and Imaging Sciences, King’s College London, London, United Kingdom, 5MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom, 6Department of Neuroimaging, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, United Kingdom, 7Radiology Department, Great Ormond Street Hospital for Children, London, United Kingdom, 8MRC Centre for Neurodevelopmental Disorders, King’s College London, London, United Kingdom |
||
The Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE) sequence is commonly used for 3D structural T1-weighted imaging of the brain at 7T and can be optimised to obtain UNI and clinically relevant FLuid and White Matter Suppression (FLAWS) images within one acquisition. In this study, such a protocol was used together with newly derived analytical equations accounting for partial Fourier acquisitions, and B1+ maps, in a dedicated fitting algorithm to produce quantitative T1 and arbitrarily scaled PD-maps. These maps were evaluated in children and adults at 7T demonstrating a significant T1 reduction with age. |
| 1245 | Computer 41
|
Meningiomas with NF2-Loss Exhibit Strong Radiomics Correlations on Contrast Enhanced T1-Weighted MRI at 3T |
| Esra Sümer1, Kübra Tan2, Ayça Ersen Danyeli3, Ozge Can4, M. Necmettin Pamir5, Alp Dinçer6, Koray Özduman5, and Esin Ozturk-Isik1 | ||
1Institute of Biomedical Engineering, Bogazici University, İstabul, Turkey, 2Health Institutes of Turkey, İstanbul, Turkey, 3Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, İstanbul, Turkey, 4Department of Medical Engineering, Acibadem Mehmet Ali Aydinlar University, İstanbul, Turkey, 5Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, İstanbul, Turkey, 6Department of Radiology, Acibadem Mehmet Ali Aydinlar University, İstanbul, Turkey |
||
Meningioma is the most common primary intracranial tumor in adults, and loss of NF2 function in meningiomas has been associated with a more aggressive biology, shorter time to recurrence and shorter overall survival. In this work, radiomics based biomarkers of NF2 copy number loss (NF2-L) have been assessed. Lower grey-level co-occurrence matrix cluster shade with wavelet high-high-low pass filter, lower original image first order minimum, and higher first order skewness with wavelet low-low-high pass filter were observed in tumors with NF2-L compared to tumors with no copy number loss. The classification accuracy of NF2 molecular subsets was 0.80±0.03 (precision=0.85±0.04, recall=0.73±0.05). |
| 1246 | Computer 42
|
Mapping Transient Co-Activity Patterns of Brain with High- and Low- Intensity Frames of fMRI Data |
| Kaiming Li1 and Xiaoping Hu1 | ||
1Department of Bioengineering, UC Riverside, Riverside, CA, United States |
||
Framewise co-activity patterns of functional MRI data may reflect the transient synchronization and coordination across brain regions. High-intensity frames of resting state fMRI have revealed granulated co-activation patterns that resembled the resting state brain networks. However, whether the low-intensity frames carry such information remains unclear. The present study trained variational autoencoder models with both positive values and negative values of normalized fMRI time series respectively and evaluated the two models on a separate dataset. We found the two models were very similar, suggesting that the negative-value frames can reflect transient brain co-activity patterns as the positive ones do. |
| 1247 | Computer 43
|
U-Net-based deep convolutional neural network for detection of superparamagnetic drug-eluting particles used for liver chemoembolization |
| NING LI1,2, Cyril Tous1,2, Phillip Fei1,2, Ivan P Dimov1,2, Simon Lessard1,2, Urs O. Häfeli3, Sylvain Martel4, An Tang1,2, and Gilles Soulez1,2 | ||
1Centre de recherche du Centre hospitalier de l’Université de Montréal (CRCHUM), Montreal, QC, Canada, 2Université de Montréal, Montreal, QC, Canada, 3University of British Columbia, Vancouver, BC, Canada, 4Polytechnique Montréal, Montreal, QC, Canada |
||
Magnetic steering of superparamagnetic drug-eluting particles (SMDEPs) loaded with anti-tumor drugs across hepatic arteries is a promising technique to perform segmental liver embolization for patients with hepatocellular carcinomas (HCCs). These aggregates (20 ± 6 SMDEPs) are sequentially released with a specially designed injector through a catheter in the proper hepatic artery of a swine placed in an MRI. Manual segmentation was previously done to localize and count the particles (volume of the artifact) in each lobe. We propose to train a U-net over this pre-existing database. Dice similarity coefficient, accuracy, and precision were respectively 99.1%, 98.3%, and 84.6%. |
| 1248 | Computer 44
|
Assessment of anal fistula using advanced diffusion-weighted imaging techniques: value of ZOOMit and RESOLVE |
| Yue Qin1, Dayong Jin1, Xin Li1, Liyao Liu1, Yinhu Zhu1, Juan Tian1, Yifan Qian1, Shaoyu Wang2, and Boyuan Jiang1 | ||
1Xi'an Daxing Hospital, Xi'an, China, 2Siemens Healthineers, Ltd., Xi'an, China |
||
This study investigated ZOOMit and RESOLVE DWI sequences to compare of the image quality of anal canal and the visibility of the fistula. The results showed that both ZOOMit DWI and RESOLVE DWI obtained high quality images， ZOOMit DWI sequence had higher fistula visibility and SNR of the fistula than those of RESOLVE DWI sequence. Our findings also suggest that ZOOMit DWI can be used as the preferred sequence for anal MR diffusion imaging. |
| 1249 | Computer 45
|
Investigation of myelin using T1/T2 ratio imaging across multiple datasets: comparable or not? |
| Riona Fumi1, Hanna Maksimuk2,3, James H Cole4,5, and Sjoerd B Vos2,4,6 | ||
1Queen Square Institute of Neurology, University College London, London, United Kingdom, 2Neuroradiological Academic Unit, Queen Square Institute of Neurology, University College London, London, United Kingdom, 3Department of Radiology, Brest Regional Clinical Hospital, Brest, Belarus, 4Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, 5Dementia Research Centre, Queen Square Institute of Neurology, University College London, London, United Kingdom, 6Centre for Microscopy, Characterisation, and Analysis, The University of Western Australia, Perth, Australia |
||
Cortical T1/T2 ratios (T1/T2r) have been used to assess myelination along the lifespan and investigate demyelination in neurological diseases. Creating a comprehensive normative population from open-access datasets could facilitate widespread adoption of this technique. Here we investigate comparability between two such publicly available datasets, OASIS3 and CamCAN. We find that they vary in terms of offset, lifespan trajectory, and presence (OASIS3) or absence (CamCAN) of hemispheric asymmetry. This variability exists despite both studies use the same scanner model, and is likely increased across different scanner models and vendors. These findings indicate that combining data from different studies requires advanced normalisation. |
| 1250 | Computer 46
|
Noise reduction in fractional anisotropy maps using deep learning based denoising |
| Seema S Bhat1, Pavan Poojar2,3, Chennagiri Rajarao Padma2, and Hanumantharaju MC4 | ||
1Department of Information Science and Engineering, Dayananda Sagar College of Engineering, Bengaluru, India, 2Department of Medical Electronics, Dayananda Sagar College of Engineering, Bengaluru, India, 3Columbia University in the City of New York, Newyork, NY, NY, United States, 4Department of Electronics and Communications, BMS Institute of Technology and Management, Bangalore, India |
||
Denoising is an alternative for enhancing signal-to-noise ratio in high b-value diffusion imaging instead of prolonged acquisition time. We experimented a deep learning based denoising method on prospective high b-value DWI and visualized the impact of denoising using fractional anisotropy(FA) maps. Experiment was repeated for three different signal averages:1,2 4-NEX and two different slice thickness 1mm and 5mm with gold standard reference of 10-NEX images. The current work obtained average peak signal-to-noise ratio >34dB and SSIM >0.94 after denoising for FA maps. The PSNR and SSIM values in FA maps were modestly increased after denoising. |
| 1251 | Computer 47
|
Resolution-dependency of arterial and venous density estimates and vessel distance maps in deep gray matter |
| Hendrik Mattern1 and Oliver Speck1,2,3,4 | ||
1Biomedical Magnetic Resonance, Otto von Guericke University, Magdeburg, Germany, 2German Center for Neurodegenerative Disease, Magdeburg, Germany, 3Center for Behavioral Brain Sciences, Magdeburg, Germany, 4Leibniz Institute for Neurobiology, Magdeburg, Germany |
||
The detection of vessel in MRI depends on the imaging resolution used. Hence, subsequent quantification with vessel densities and vessel distance mapping (VDM) is resolution dependent. In this study, the resolution dependency of both metrics was evaluated for arteries and veins in deep gray matter regions by retrospective downsampling of high-resolution vessel images. Quantitative comparison shows that with decreasing resolution the estimated vessel distances increase, while vessel densities decline. The resolution dependency of vessel densities and distances can be modeled linearly. The here found scaling factors may improve inter-study comparability of vessel densities and VDM. |
| 1252 | Computer 48
|
MRCP at 1.5 T: comparing image quality and acquisition time between respiratory-triggered and breath-hold SPACE sequences |
| Xin Li1, Yue Qin1, Dayong Jin1, Yinhu Zhu1, Liyao Liu1, Yifan Qian1, Juan Tian1, and Shaoyu Wang2 | ||
1Xi'an Daxing Hospital, Xi'an, China, 2Siemens Healthineers, Ltd., Xi'an, China |
||
This study used magnetic resonance cholangiopancreatography (MRCP) to compare the image quality and acquisition time of conventional respiratory-triggered (RT), modified RT, and breath-hold (BH) sampling perfection with application-optimized contrasts using different flip evolutions (SPACE) sequences in patients with biliary and pancreatic disorders. Our results suggested that modified RT-SPACE-MRCP sequence can significantly reduced acquisition time and had better image quality. The MRCP sequence can be selected flexibly according to the patient's situation to obtain good diagnostic images. |
| 1253 | Computer 49
|
Quantifying metabolites in large brain areas by 3D Echo Planar Spectroscopic Imaging spectral averaging for a study of cigarette smoking |
| Jeffry R Alger1,2,3,4, Maylen Perez Diaz5, Joseph O'Neill5, Dara Ghahremani5, Andy C Dean5, Sulaiman Sheriff6, Andrew A Maudsley6, and Edythe London5 | ||
1NeuroSpectroScopics LLC, Sherman Oaks, CA, United States, 2University of California, Los Angeles, Los Angeles, CA, United States, 3Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, TX, United States, 4Hura Imaging Inc, Calabasas, CA, United States, 5Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, Los Angeles, CA, United States, 6Radiology, University of Miami, Miami, FL, United States |
||
Spectral averaging of thousands of magnetic resonance spectroscopic imaging voxel spectra over large brain volumes provides a means of assessing low-level and/or hard-to-quantify metabolite differences over large brain volumes between groups of individuals. |
| 1254 | Computer 50
|
Different Brain Areas Require Different Analysis Models—fMRI Observations in Parkinson’s Disease. |
| Renzo Torrecuso1, Karsten Mueller1, Stefan Holiga1,2, Dusan Urgosik3, Josef Vymazal4, Tomas Sieger5, Filip Růžička6, Evzen Růžička7, Matthias Schroeter8, Robert Jech3, and Harald E. Möller1 | ||
1NMR, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland, 3Department of Neurology, Charles University in Prague | CUNI, Prague, Czech Republic, 4| CULS · Faculty of Environmental Science, Czech University of Life Sciences Prague, Prague, Czech Republic, 5Department of Neurology and Center of Clinical Neuroscience, Charles University in Prague, Prague, Czech Republic, 6Department of Environmental Engineering, Faculty of Technology Tomas Bata University in Zlín, Zlin, Czech Republic, 7Neurology, General University Hospital in Prague, Prague, Czech Republic, 8Clinic for Cognitive Neurology, University Hospital Leipzig, Leipzig, Germany |
||
Foreseeing how specific brain areas respond in time to a stimulus can be a prerequisite for a successfully conceived fMRI experiment. We demonstrate that in medicated Parkinson’s disease patients, putamen's activation peaks around the onset of tapping but does not persist throughout the tapping block, whereas sustained activation is observed in the motor cortex. Consequently, in the widely used tapping paradigm “On vs. Off L-DOPA”, the drug effect remains undetected if statistical analysis apply a block design instead of an event-related one. Ignoring this information can lead to fallacious conclusions which suggests using different models to investigate different brain regions. |
| 1255 | Computer 51
|
Test-retest reliability of velocity pulsatility in perforating arteries in the basal ganglia at 3T MRI |
| Rick J. van Tuijl1, Stanley D.T. Pham1, Birgitta K Velthuis1, Ynte M. Ruigrok2, Geert Jan Biessels2, and Jaco J.M. Zwanenburg1 | ||
1Radiology, UMC Utrecht, Utrecht, Netherlands, 2Neurology, UMC Utrecht, Utrecht, Netherlands |
||
We studied the test-retest reliability of assessing perforating arteries of the BG with 2D-PC velocity measurements at 3T-MRI, with and without repositioning of the patient. We showed reproducible assessment of Ndetected, Vmean and vPI in the perforating arteries, independent of slice planning. The relative imprecision was smallest for Vmean (12%) and highest for vPI (32%). The interrater-reliability assessment showed high consistency between two raters (mean variability <5% for all parameters). We found very similar Limits of Agreement with and without repositioning, which indicated that uncertainty in measuring these parameters was dominated by scanner and physiological noise, rather than by planning. |
| 1256 | Computer 52
|
T2-dependence of deep and superficial white matter tractography |
| Yogesh Rathi1, Lipeng Ning1, Congyu Liao2, Yang Ji3, Carl-Fredrik Westin1, Fan Zhang1, Nikolaos Makris1, Lauren J O'Donnell1, and Kawin Setsompop2 | ||
1Harvard Medical School, Boston, MA, United States, 2Stanford University, Stanford, CA, United States, 3Oxford University, Oxford, United Kingdom |
||
Tractography is typically obtained from diffusion MRI data acquired at a particular echo time (TE). Recent works have shown differential T2-relaxation times of tissue microstructure and fiber tracts. However, estimation of T2 in fiber bundles have typically been done at the last stage (after tractography). In this work, we show that tractography itself is affected by the T2 (and hence TE) of the fibers. Specifically, we show that the ability to trace accurately depends on the T2 of the fiber bundle. This is specifically true for the superficial white matter (u-fibers) which has higher myelination and iron content. |
| 1257 | Computer 53
|
Impact of Diffusion Signal Harmonisation on Voxel-Wise Analysis of Mild Traumatic Brain Injury |
| Stefan Winzeck1,2, Sophie Richter2, Maira Siqueira Pinto3, Evgenios N. Kornaropoulos4, Virginia F.J. Newcombe2, Ben Glocker1, David K. Menon2, and Marta M. Correia5 | ||
1BioMedIA, Department of Computing, Imperial College London, London, United Kingdom, 2University Division of Anaesthesia, Department of Medicine, University of Cambridge, Cambridge, United Kingdom, 3Universitair Ziekenhuis Antwerpen, Antwerp, Belgium, 4Clinical Sciences, Lund, Diagnostic Radiology, Lund University, Lund, Sweden, 5MRC Cognition and Brain Sciences Unit, University of Cambridge, Cambridge, United Kingdom |
||
The comparison of voxel-wise regression analysis of a multi-centre diffusion MRI study showed that similar results could be achieved regardless of whether prior data harmonisation was applied or not. However, harmonisation had a positive impact on increasing the effect size found between mild traumatic brain injury patients and controls. |
| 1258 | Computer 54
|
Combining Shape Descriptor and Image Ranking to the Identification of Alzheimer’s Disease in Anatomical Regions using T1-weighted MR Images |
| Kaue Tartarotti Nepomuceno Duarte1, Richard Frayne1, David Gobbi2, Paulo Sergio Martins3, and Marco Antonio Garcia de Carvalho3 | ||
1Radiology, University of Calgary, Calgary, AB, Canada, 2CIPAC, University of Calgary, CALGARY, AB, Canada, 3University of Campinas, Limeira, Brazil |
||
Alzheimer's Disease is one of the most prevalent types of dementia that affects human behaviour and cognitive skills. The progression of this dementia occurs at different rates in distinct parts of the brain, thus also affecting their brain shape. We proposed an approach based on shape feature extraction and image ranking to identify which regions are more predictive for each level of the AD progression. For each stage, we highlighted the sub-cortical regions that suggested a strong correlation in the AD levels. In fact, the results showed a predictive pattern, which aligns to the state-of-the-art. |
| 1259 | Computer 55
|
Determinants of MR Relaxation in the Extremely Preterm Brain at Adolescence: Myelin and Iron |
| Ryan McNaughton1,2, Ning Hua2, Lei Zhang3, David Kennedy4, T. Michael O'Shea3, Karl Kuban2, and Hernan Jara2 | ||
1Mechanical Engineering, Boston University, Boston, MA, United States, 2Boston University Medical Center, Boston, MA, United States, 3University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 4University of Massachusetts, Worcester, MA, United States |
||
Purpose: To describe the dependencies of multiparametric quantitative MRI (MP-qMRI) on myelin and iron content in the extremely preterm born brain at adolescence. Methods: Algorithms for a fast exchange relaxation (FER) model and MP-qMRI create maps of R1, R2, myelin, and iron in 30 participants using MR images obtained with the triple TSE pulse sequence at age 15 years. Results: R1 and R2 have linear dependencies with myelin and iron content, respectively. Conclusion: Application of a FER model produces coregistered maps of myelin and iron content which exhibit unique influences on the relaxation of white matter and gray matter regions. |
| 1260 | Computer 56
|
Improved ΔR2* calculation through voxelwise subtraction for MRI-based dosimetry of holmium-166 transarterial radioembolization |
| Meike W.M. van Wijk1, Joey Roosen1, Lovisa E.L. Westlund Gotby1, Mark J. Arntz1, Marcel J.R. Janssen1, Daphne Lobeek1, Gerrit H. van de Maat2, Christiaan G. Overduin1, and J. Frank W. Nijsen1 | ||
1Department of Medical Imaging, Radboud University Medical Center, Nijmegen, Netherlands, 2Imaging & Software Solutions, Quirem Medical B.V., Deventer, Netherlands |
||
Transarterial radioembolization (TARE) is a treatment for liver cancer, during which radioactive microspheres are administered through the hepatic artery. Microspheres containing holmium-166 enable MRI-based dosimetry, based on subtraction of pre- and post-treatment $$$R_2^*$$$ values. This subtraction is performed using a mean pre-treatment $$$R_2^*$$$ value. This does however not take pre-existing differences of $$$R_2^*$$$ values into account, introducing an error in the dosimetry. In this work a voxelwise subtraction method is presented, using deformable registration to transform the pre-treatment $$$R_2^*$$$ map to the post-treatment $$$R_2^*$$$ map, enabling voxel-by-voxel subtraction. This method does take $$$R_2^*$$$ differences into account and improves MRI-based dosimetry. |
| 1261 | Computer 57
|
The Compressed Sensing MP2RAGE as a surrogate to the MPRAGE for neuro-imaging at 3T |
| Aurélien J Trotier1, Bixente Dilharreguy2, Serge Anandra3, Nadège Corbin1, William Lefrançois1, Valery Ozenne1, Sylvain Miraux1, and Emeline Julie Ribot1 | ||
1Centre de Résonance Magnétique des Systèmes Biologiques, UMR 5536, CNRS, Bordeaux, France, 2UMS3767, CNRS, Bordeaux, France, 3UMS3767, Université de Bordeaux, Bordeaux, France |
||
A Compressed Sensing Magnetization prepared Two Rapid Gradient Echo (CS-MP2RAGE) sequence was developed to provide images in <10min including reconstruction time. It was applied on 13 participants each scanned 3 to 4 times with repositioning. Compared to the standard MPRAGE sequence, it provides higher contrast morphological images. Similar intra-volunteer variabilities in volume segmentations of the brain structures were obtained. Additionally, high resolution T1 maps provided T1 values of white and gray matters and several deep grey nuclei consistent with the literature, and show very low variability (<1%). The CS-MP2RAGE can be used in future clinical research protocols. |
| 1262 | Computer 58
|
Whole brain 3D radial T2 mapping with retrospective motion correction capabilities. |
| Nadège Corbin1,2, Aurélien J Trotier1, Sylvain Miraux1, and Emeline J Ribot1 | ||
1Centre de Résonance Magnétique des Systèmes Biologiques, UMR 5536, Bordeaux, France, 2Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom |
||
T2 mapping is usually very time consuming and long acquisition time often result in high sensitivity to intra-scan motion. Here we propose a T2 mapping method relying on 3D radial sampling and advanced iterative reconstruction to retrospectively correct for motion and/or reduce the acquisition. First, in-vivo experiments at 3T showed the capability of the technique to correct for intra-scan motion and therefore improve the quality of resulting T2 maps. Second, acquisitions varying from 7 to 33 min of whole-brain single and multi-compartment T2 maps were obtained, showing the potential of the method to accelerate and therefore reduce sensitivity to motion. |
| 1263 | Computer 59
|
Improving T1 mapping with a golden-angle radial MOLLI and a model-based regularized reconstruction SALSA-EPG |
| Andreia C Freitas1, Andreia S Gaspar1, Nuno A Silva2, José M Bioucas-Dias3, and Rita G Nunes1 | ||
1ISR Lisboa/LARSyS and Department of Bioengineering, Instituto Superior Técnico - Universidade de Lisboa, Lisbon, Portugal, 2Hospital da Luz Learning Health, Luz Saúde, Lisbon, Portugal, 3Instituto de Telecomunicações, Instituto Superior Técnico - Universidade de Lisboa, Lisbon, Portugal |
||
T1 mapping provides valuable information regarding cardiovascular pathologies. Clinical approach consists of using a Cartesian MOLLI sequence and fitting a 3-parameter model. Although easy and straightforward, this method can require long breath-holds and does not explicitly include other factors affecting T1 estimation. We propose coupling an accelerated golden-angle radial MOLLI with a model-based regularized reconstruction (SALSA). The proposed method was tested in phantom and in vivo data at 1.5T. Improved T1 precision and good accuracy was found for the phantom data. Lower T1 was estimated with SALSA compared to the commercial sequence in vivo, future work will address this. |
| 1264 | Computer 60
|
Robust and efficient R2* estimation in human brain using log-linear weighted least squares |
| Luke J. Edwards1, Siawoosh Mohammadi1,2, Kerrin J. Pine1, Martina F. Callaghan3, and Nikolaus Weiskopf1,4 | ||
1Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, 3Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 4Felix Bloch Institute for Solid State Physics, Faculty of Physics and Earth Sciences, Leipzig University, Leipzig, Germany |
||
In vivo maps of R2* in human brain hold promise for neuroscientific investigations. We tested several R2* fitting routines (nonlinear least squares [NLLS; silver standard], auto-regression on linear operations [ARLO], log-linear weighted least square [WLS], and log-linear ordinary least squares [OLS]) for dual flip angle multi-echo FLASH data in simulations and challenging 400 µm resolution in vivo 7T data. Log-linear WLS was found to give a good trade-off between accuracy, precision, and computational time. The method will be available in a future version of the open source hMRI toolbox (hmri.info). |
| 1265 | Computer 61
|
Quantitative Relaxometry Metrics for Normal and Brain Tumor Tissues using MR Fingerprinting and Magnetic Resonance Image Compilation Methods |
| Amaresha Shridhar Konar1, Akash Deelip Shah2, Abhay Dave3, Suchandrima Banerjee4, Maggie Fung5, Vaios Hatzoglou2, and Amita Shukla-Dave1,2 | ||
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York City, NY, United States, 2Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States, 3Touro College of Osteopathic Medicine, New York, NY, United States, 4GE Healthcare, New York, NY, United States, 5GE Healthcare, New York City, NY, United States |
||
This work aims to assess the relaxometry maps generated using MRF and MAGiC in brain tumor (n=27). A total of 14 brain tumor tissue regions (Pre-Tx n=5, Post-Tx n=9) were used for comparison. The T1 and T2 values measured at tumor and normal appearing (contralateral) region showed significant difference in both MAGiC and MRF generated maps. The results show that the relaxometry values estimated using MRF and MAGiC methods can differentiate Tumor and normal appearing tissues. The utility of this technique needs to be further explored in larger sample studies and it could be further extendable to classify tumor types. |
| 1266 | Computer 62
|
Feasibility of fast whole brain T2 mapping using CAIPIRINHA accelerated SPACE sequences |
| Christoph Birkl1, Christian Kremser2, Alexander Rauscher3, and Elke Ruth Gizewski1 | ||
1Department of Neuroradiology, Medical University of Innsbruck, Innsbruck, Austria, 2Department of Radiology, Medical University of Innsbruck, Innsbruck, Austria, 3UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada |
||
We investigated whether multiple SPACE sequences accelerated using CAIPIRINHA and acquired with different echo times can be used to obtain high resolution isotropic whole brain T2 maps within a clinical feasible scan time. In a phantom study, we could show that the SPACE based T2 values are comparable with T2 values derived using multiple single spin echo sequences. By reducing the number of echoes we could acquire 1 mm isotropic in vivo whole brain T2 maps in under 3 minutes. |
| 1267 | Computer 63
|
Magnetisation transfer saturation (MTsat) and MTR: relationship with T1 recovery in multiple sclerosis and healthy brain |
| Elizabeth N. York1, Rozanna Meijboom1, Agniete Kampaite1, Maria Valdes Hernandez1, Michael J. Thrippleton1, and Adam D. Waldman1 | ||
1Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom |
||
Magnetisation transfer saturation (MTsat) shows improved tissue contrast compared with magnetisation transfer ratio (MTR), due to T1 correction, and is more sensitive to subtle myelin loss in multiple sclerosis (MS). It is not clear how MTsat performs in severely demyelinated tissue with markedly increased T1. We examine the relationship between T1app, MTsat and MTR in healthy and MS brains. We show a negative linear relationship between MTsat and T1app in white matter, which breaks down in grey matter and white matter lesions. MTsat is sensitive to modest myelin disruption, while T1app may better reflect severe damage in MS lesions. |
| 1268 | Computer 64
|
High-resolution T1 atlas for subject-specific abnormality detection at 7T |
| Gian Franco Piredda1,2,3, Piotr Radojewski4,5, Arun Joseph5,6,7, Gabriele Bonanno5,6,7, Karl Egger8, Shan Yang8, Punith B. Venkategowda9, Ricardo A. Corredor-Jerez1,2,3, Bénédicte Maréchal1,2,3, Roland Wiest4,5, Jean-Philippe Thiran2,3, Tom Hilbert1,2,3, and Tobias Kober1,2,3 | ||
1Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, 2Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 3LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, 4Support Center for Advanced Neuroimaging, Institute for Diagnostic and Interventional Neuroradiology, Inselspital, Bern University, Bern, Switzerland, 5Translational Imaging Center, sitem-insel AG, Bern, Switzerland, 6Advanced Clinical Imaging Technology, Siemens Healthcare AG, Bern, Switzerland, 7Magnetic Resonance Methodology, Institute of Diagnostic and Interventional Neuroradiology, University of Bern, Bern, Switzerland, 8Department of Neuroradiology, Medical Center – University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 9Siemens Healthcare Pvt. Ltd., Bangalore, India |
||
Previous studies at 3T have shown that T1 relaxometry enables personalized characterization of brain tissues by comparing physical properties of a single patient to a normative atlas. Ultra-high field imaging allows exploiting this concept at even higher resolutions, which can be crucial to detect certain diseases. To this end, here we established an atlas of normative T1 values at 7T from acquisitions with 0.6$$$\times$$$0.6$$$\times$$$0.6 mm3 isotropic resolution. Additionally, the clinical potential and improvement of 7T vs. 3T imaging is shown in two case reports from patients scanned at both field strengths. |
| 1269 | Computer 65
|
TR dependence of phase-cycled bSSFP relaxometry in brain tissue |
| Jessica Schäper1,2 and Oliver Bieri1,2 | ||
1Department of Biomedical Engineering, University of Basel, Basel, Switzerland, 2Division of Radioligical Physics, Department of Radiology, University of Basel Hospital, Basel, Switzerland |
||
Brain relaxometry with phase-cycled bSSFP shows systematically lower T1 values, if compared to spoiled-GRE or inversion-recovery spin echo methods. One explanation can be the pronounced asymmetry in the bSSFP's frequency profile, observed for tissues. It was recently shown that this asymmetry decreases towards shorter TR, possibly leading to an adjustment of T1 estimates from bSSFP to spoiled-GRE. Here, it was investigated how T1 and T2 quantification is influenced by TR. Contrary to expectation, a stronger mismatch between bSSFP and spoiled-GRE was observed towards shorter TR. The origin of this mismatch can thus not be attributed to the bSSFP profile asymmetry. |
| 1270 | Computer 66
|
A comprehensive protocol for multiparametric brain MRI |
| Dvir Radunsky1, Chen Solomon1, Tamar Blumenfeld-Katzir1, Neta Stern1, Shir Filo2, Aviv Mezer2, Anita Karsa3, Karin Shmueli3, Lucas Soustelle 4, Guillaume Duhamel4, Olivier M. Girard4, and Noam Ben-Eliezer1,5,6 | ||
1Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel, 2The Edmond and Lily Safra Center for Brain Sciences, The Hebrew University of Jerusalem, Jerusalem, Israel, 3Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 4Aix Marseille University, CNRS, CRMBM, Marseille, France, 5Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel, 6Center for Advanced Imaging Innovation and Research (CAI2R), University Langone Medical Center, New York, Israel |
||
The clinical utility of quantitative MRI (qMRI) techniques was demonstrated in numerous pathologies. This work investigated a range of qMRI pulse-sequences and processing methods with proven clinical applicability, aiming to establish a comprehensive and standard qMRI scan protocol for the brain. This multiparametric protocol provides a wide range of numeric maps (e.g., T1, T2, T2*, PD, M0, B1+, water and macromolecular fractions, susceptibility, mean diffusivity, and more) with whole-brain coverage, diverse set of clinical biomarkers, and the ability for stable longitudinal and multi-center investigations. Limitations and practical tips are provided for users interested in quantitative brain imaging. |
| 1271 | Computer 67
|
Improved spatial normalization of white matter fiber orientation distributions using T1-weighted contrast |
| Jose M Guerrero-Gonzalez1,2, Olivia Surgent2,3, Nagesh Adluru2,4, Steven R Kecskemeti2, Gregory R Kirk2, Douglas C Dean III1,2,5, Brittany G Travers2,6, and Andrew L Alexander1,2,7 | ||
1Medical Physics, University of Wisconsin - Madison, Madison, WI, United States, 2Waisman Center, University of Wisconsin - Madison, Madison, WI, United States, 3Neuroscience Training Program, University of Wisconsin - Madison, Madison, WI, United States, 4Radiology, University of Wisconsin - Madison, Madison, WI, United States, 5Pediatrics, University of Wisconsin - Madison, Madison, WI, United States, 6Kinesiology Occupational Therapy Program, University of Wisconsin - Madison, Madison, WI, United States, 7Psychiatry, University of Wisconsin - Madison, Madison, WI, United States |
||
Fiber orientation distributions (FOD) derived from diffusion magnetic resonance imaging (dMRI) enable resolution of multiple fiber populations within a voxel. FOD-based white matter studies include voxel-based analysis, atlas-based labeling, and group average fiber tracking. These methods require spatial normalization of the FODs. This work describes an alternative approach for FOD spatial normalization based on co-registering individual dMRI to the T1-weighted (T1w) images, non-linear spatial normalization of the T1w images to a template, and applying the transformations to the FOD maps. This approach is compared to the conventional approach of directly aligning FOD maps. |
| 1272 | Computer 68
|
Disentangling the contributions of myelin, neurofilament & microglia to MR contrast: an automated pipeline for voxelwise MR-histology analysis |
| Daniel Z.L. Kor1, Saad Jbabdi1, Jeroen Mollink1, Istvan N. Huszar1, Sean Foxley2, Menuka Pallebage-Gamarallage3, Connor Scott3, Adele Smart3, Olaf Ansorge3, Karla L. Miller1, and Amy F.D. Howard1 | ||
1Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, 2Department of Radiology, University of Chicago, Chicago, IL, United States, 3Neuropathology, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom |
||
Acquisition of MRI and histology in the same ex-vivo tissue sample enables direct correlation between MR and histologically-derived metrics. Here, we analysed immunohistochemistry images of human visual cortex, anterior cingulate and hippocampus to produce stained area fraction maps for myelin, neurofilaments and microglia. We performed voxelwise correlations between MR parameters (FA, MD, R2*, R1) and histology maps to generally characterise the strength of relationships. We then used partial correlation to identify the unique variance in MR parameters explained by each histological feature, and multiple linear regression to explore how well multiple microstructural properties can together explain MR parameters. |
| 1273 | Computer 69
|
End-to-End Scan Parameter Optimization for Improved Myelin Water Imaging |
| Steven T. Whitaker1, Jon-Fredrik Nielsen2, and Jeffrey A. Fessler1 | ||
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States, 2Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States |
||
Optimization techniques can be used to design scan parameters for quantitative imaging. The Cramér-Rao Lower Bound (CRLB) is often used for such designs, but it only characterizes unbiased estimators. We propose an end-to-end approach to scan design that optimizes scan parameters with a particular estimator in mind. We compare CRLB-based and end-to-end scan designs in the context of myelin water imaging. The end-to-end scan design results in lower estimation error in simulation and an in vivo myelin water fraction (MWF) map with improved contrast. The proposed end-to-end scan design approach is thus a promising alternative to using the CRLB. |
| 1329 | Computer 35
|
Radiomic biomarker extracted from PI-RADS 3 patients support more efficient and robust prostate cancer diagnosis: a multi-center study |
| Longfei Li1,2, Rui Yang3, Xin Chen4, Cheng Li2, Hairong Zheng2, Yusong Lin1, Zaiyi Liu4, and Shanshan Wang2 | ||
1the Collaborative Innovation Center for Internet Healthcare , School of Information Engineering, Zhengzhou University, Zhengzhou, China, 2Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, 3Department of Urology, Renmin Hospital of Wuhan University, Wuhan, China, 4Department of Radiology, Guangdong Provincial People’s Hospital, Guangzhou, China |
||
Prostate Imaging Reporting and Data System (PI-RADS) based on multi-parametric MRI classifies patients into 5 categories (PI-RADS 1-5) for routine clinical diagnosis guidance. However, there is no consensus on whether PI-RADS 3 patients should go through biopsies. Mining features from these hard samples (HS) is meaningful for physicians to achieve accurate diagnoses. Currently, the mining of HS biomarkers is insufficient, and the effectiveness and robustness of HS biomarkers for prostate cancer diagnosis have not been explored. In this study, biomarkers from different data distributions are constructed. Results show that HS biomarkers can achieve better performances in different data distributions. |
| 1330 | Computer 36
|
Workflow development for kidney segmentation using a U-net model using 11000 MRI data sets from the German National Cohort (NAKO/GNC) study |
| Martin Buechert1, Jan Lipovsek2, Marco Reisert2, Harald Horbach2, Wilfried Reichardt2, Christopher Schlett2, Fabian Bamberg2, Peggy Sekula 2, Anna Köttgen2, and Elias Kellner2 | ||
1Magnetic Resonance Development and Applicatione Center, Medical Center, University of Freiburg, Faculty of Medicine, Freiburg, Germany, 2Medical Center, University of Freiburg, Faculty of Medicine, Freiburg, Germany |
||
A processing pipeline for kidney segmentation using a hierarchical patch-based stack of U-nets was implemented and applied to abdominal MRI images of the German National Cohort study. The training data set included 300 cases and the final net was applied to the dataset of 11,207 MRIs. The compartments cortex, medulla and hilus could be segmented very robustly with the network. The relation of first parameters based on the segmentation withsex, age, weight, subject size and BMI are presented. This is an optimal starting point to identify more advanced biomarkers and their correlations, especially with kidney functional parameters. |
| 1331 | Computer 37
|
Assessing Inter-Observer Variability of MRI Tagging of the Colon Contents in Healthy Human Subjects |
| Meshari T Alshammari1,2, Ali S Alyami3, Victoria Wilkinson-Smith3, Luca Marciani1, Gordon W. Moran1, and Caroline L. Hoad3,4 | ||
1Translational Medical Sciences and National Institute for Health Research (NIHR) Nottingham Biomedical Research Centre, Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, United Kingdom, 2Department of Diagnostic Radiology, College of Applied Medical Sciences, University of Hail, Hail, Saudi Arabia, 3NIHR Nottingham Biomedical Research Centre (BRC), Nottingham University Hospitals NHS Trust and University of Nottingham, Nottingham, United Kingdom, 4Sir Peter Mansfield Imaging Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, United Kingdom |
||
MRI Tagging techniques have been applied to the GI tract to assess bowel contractions and colon content mixing. Two independent datasets of healthy adults were used to evaluate the dependence of the percentage Coefficient of variation (%CoV) measurement on inter-observer variability in the ascending colon and descending colon. There was little difference between both AC and DC measurements of the %CoV with high inter-rater agreement (intra-class correlation coefficient > 0.95). This technique could be used to provide objective measures for the motility assessment of the colon in inflammatory and functional bowel diseases.
|
| 1332 | Computer 38
|
Optimization of GLCM Texture Analysis settings in Liver Radial T2 Maps: Fibrotic vs Healthy Liver |
| Diana Bencikova1,2, Veronika Janacova1, Marcus Raudner1, Ahmed Ba-Ssalamah1, Siegfried Trattnig1,2, and Martin Krssak3 | ||
1Department of Biomedical Imaging and Image-Guided Therapy, Medical University Vienna, Vienna, Austria, 2Christian Doppler Laboratory for Clinical Molecular Imaging, MOLIMA, Vienna, Austria, 3Division of Endocrinology and Metabolism, Department of Medicine III, Medical University Vienna, Vienna, Austria |
||
GLCM texture analysis is a promising technique for characterizing and classifying tissue pathologies. In liver applications it has been mostly performed on T2-weighted images. Fast radial T2 mapping techniques enable acquiring T2 maps in clinically feasible measurement time (during breath-hold). Here, we explored the performance of the different settings of GLCM texture analysis in liver T2 maps, which might be more advantageous compared to T2-weighted images. We identified the grey-level quantization of 8bits and direction of 90° as the best setting for discrimination between fibrotic and healthy livers. |
| 1333 | Computer 39
|
Analyzing Radiomic Features in Magnetic Resonance Images of Head and Neck Cancer during Radiation Therapy: Preliminary Results |
| Victor Fritz1, Martin Schwartz1,2, Jens Kübler3, Simon Böke4, Jonas Habrich5, Daniel Zips4, Daniela Thorwarth5, Konstantin Nikolaou3, and Fritz Schick1 | ||
1Department of Diagnostic and Interventional Radiology, University of Tuebingen, Section on Experimental Radiology, Tuebingen, Germany, 2University of Stuttgart, Institute of Signal Processing and System Theory, Stuttgart, Germany, 3Department of Diagnostic and Interventional Radiology, University of Tuebingen, University Hospital Tuebingen, Tuebingen, Germany, 4Department of Radiation Oncology, University of Tuebingen, University Hospital and Medical Faculty, Tuebingen, Germany, 5Department of Radiation Oncology, University of Tuebingen, Section for Biomedical Physics, Tuebingen, Germany |
||
The aim of this work was to identify potential texture features as imaging biomarkers for monitoring treatment response in head and neck cancer (HNC). For this purpose, a total number of 93 texture features were extracted on segmented calculated ADC maps and compared at baseline and in the early treatment phase of radiation therapy (RT). Fifteen texture features showed a statistically difference in the course of RT. In particular, features suggesting that ADC-based HNC texture became finer but more heterogeneous changed significantly. Presented preliminary results offer initial findings that will be systematically investigated in upcoming studies. |
| 1334 | Computer 40
|
Developing a Delta Radiomics Framework for Prostate Cancer Progression Biomarkers in Patients under Active Surveillance: Pilot Study |
| Dyah Ekashanti Octorina Dewi1, Mohammed R. S. Sunoqrot1, Gabriel Addio Nketiah1, Elise Sandsmark2, Sverre Langørgen2, Helena Bertilsson3, Mattijs Elschot1,2, and Tone Frost Bathen1,2 | ||
1Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway, 2Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway, 3Department of Urology, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway |
||
Identification of progressive lesions over time is important to monitor the prostate cancer progression in active surveillance. Delta Radiomics, computed as the relative changes of features among scans, provides temporal evolution of features that may indicate clinical changes along time periods. This pilot study aims to develop a framework to identify progressive lesions on two long-term scans in prostate T2WI based on Delta-radiomics. The current results show that four Delta-radiomics features from GLDM and Shape feature groups have potentials to identify progressive lesions compared to other selected reproducible features. |
| 1335 | Computer 41
|
Quantitative analysis of dynamic-contrast enhanced MRI by dispersion analysis for improved prostate cancer diagnosis |
| Simona Turco1, Catarina Dinis Fernandes2, Razvan Miclea3, Ivo Schoots4, Peet Nooijen5, Hans van der Linden5, Jelle Barentsz6, Stijn Heijmink 7, Hessel Wijkstra2, and Massimo Mischi2 | ||
1Electrical Engineering, Eindhoven Univeristy of Technology, Eindhoven, Netherlands, 2Electrical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands, 3Abdominal radiology, Maastricht University Medical Center, Maastricht, Netherlands, 4Abdominal radiology, Erasmus Medical Center, Rotterdam, Netherlands, 5Uropathology, Jeroen Bosch Ziekenhuis, 's-Hertogenbosch, Netherlands, 6Radiology, Radboud university medical center, Nijmegen, Netherlands, 7Radiology, Netherlands Cancer Institute, Amsterdam, Netherlands |
||
Multiparametric MRI, including T2-weighted, diffusion-weighted, and dynamic contrast enhanced (DCE) imaging, is the recommended imaging modality for prostate cancer (PCa). Although the role of DCE-MRI has been greatly limited in recent years, only qualitative analysis of DCE is currently performed. Here we propose magnetic resonance dispersion imaging (MRDI) to obtain quantitative parametric maps from DCE-MRI. Comparing the performance of two radiologists, our results show that some clinically-significant PCa are only found by interpretation of either mpMRI alone or MRDI maps alone, suggesting that combined interpretation of MRDI and mpMRI may improve PCa diagnosis. |
| 1336 | Computer 42
|
Comparison of magnetization transfer quantification methods in the sciatic nerve of Charcot-Marie-Tooth Type 1A patients |
| Alison Roth1, Yongsheng Chen2, Jun Li2,3, and Richard D. Dortch1,4,5 | ||
1Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, 2Department of Neurology, Wayne State University, Detroit, MI, United States, 3Department of Neurology, Vanderbilt University, Nashville, TN, United States, 4Vanderbilt University Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States, 5Department of Biomedical Engineering, Vanderbilt University, Nashville, TN, United States |
||
Magnetization transfer (MT) can be quantified as the MT ratio (MTR), MT saturation (MTsat), and quantitative MT (qMT) pool-size-ratio (PSR). MTR is a promising peripheral nerve imaging biomarker in inherited neuropathies but has lower pathological specificity compared to MTsat and PSR. Here, MTR and MTsat metrics are compared to PSR. Eighteen subjects received 16 qMT scans, five T1-weighted scans, and the B0 and B1 fields were measured. Mean MTR, MTsat, and PSR were extracted from the sciatic nerve. MTR had the highest SNR (median>15.76) and scan-rescan repeatability (ICC=0.93, CV=4.6%), whereas MTsat had the strongest correlation to PSR (r=0.94, p<0.001). |
| 1337 | Computer 43
|
Post mortem study of R2* and vessel distance maps across cortical depth |
| Hendrik Mattern1, Frank Angenstein2, Christian Mawrin3, and Valentina Perosa2,4,5 | ||
1Biomedical Magnetic Resonance, Otto von Guericke University, Magdeburg, Germany, 2German Center for Neurodegenerative Disease, Magdeburg, Germany, 3Institute of Neuropathology, Medical Faculty at the Otto von Guericke University, Magdeburg, Germany, 4J. Philip Kistler Stroke Research Center, Massachusetts General Hospital, Boston, MA, United States, 5Department of Neurology, Otto-von-Guericke University, Magdeburg, Germany |
||
In this proof-of-principle study, an image acquisition and processing pipeline was developed to assess the cortical vasculature and its surrounding tissue. To that end, a block of human brain, which included cortex and juxtacortical white matter, was scanned post-mortem at high resolution with a multi-echo GRE sequence at 9.4T. Subsequently, R2* and vessel distance maps (VDM) were computed and their layer-specific patterns analyzed. |
| 1338 | Computer 44
|
Free-Breathing R2* Mapping for Hepatic Iron Quantification Using Respiratory Motion-Resolved 3D Multi-Echo UTE Cones MRI |
| MungSoo Kang1,2, Yan Wen3, Michael Carl3, Gerald G. Behr4, Ricardo Otazo1,4, and Youngwook Kee1 | ||
1Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2Department of Biomedical Engineering, Ulsan National Institute of Science and Technology, Ulsan, Korea, Republic of, 3GE Healthcare, Waukesha, WI, United States, 4Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States |
||
Respiratory motion is one of the major factors hindering accurate R2* mapping for hepatic iron quantification. In this work, motion-resolved 3D multi-echo UTE cones MRI with pseudo-random view ordering was implemented for motion-robust and free-breathing R2* mapping of hepatic iron. Compared to conventional gridding reconstruction, motion-resolved reconstruction reduced the overestimation of R2* induced by respiratory motion artifacts, enabling an accurate measurement of R2*. 3D multi-echo UTE cones MRI with motion-resolved reconstruction demonstrated the feasibility of accurate free-breathing R2* mapping for hepatic iron quantification. |
| 1339 | Computer 45
|
A Recurrent Encoder-Decoder Network Accelerates T2 Mapping in Knee, Hip and Lumbar Spine |
| Aniket Tolpadi1,2, Francesco Calivà1, Misung Han1, Valentina Pedoia1, and Sharmila Majumdar1 | ||
1Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States, 2Bioengineering, University of California, Berkeley, Berkeley, CA, United States |
||
Compositional MRI (cMRI) offers sensitivity to biochemical changes that can precede morphological changes in tissues and offer quantitative assessments of musculoskeletal tissue health but suffer from long scan times. We present a recurrent encoder-decoder architecture that predicts ground truth T2 maps from spatially undersampled echo images. Reconstructed maps showed strong fidelity to ground truth in lumbar spine IVDs and knee and hip knee cartilage. Reconstruction errors were below clinically significant T2 changes through R=12 in the knee and lumbar spine and R=8 in the hip. This work marks progress towards a clinical pipeline that reduces cMRI acquisition time. |
| 1340 | Computer 46
|
Open-Source Myocardial T1 mapping accelerated with SMS: combining an auto-calibrated blip-bSSFP readout with VERSE-MB pulses |
| Andreia S Gaspar1, Nuno A Silva2, and Rita G Nunes1 | ||
1Institute for Systems and Robotics - Lisboa and Department of Bioengineering, Instituto Superior Técnico, Universidade de Lisboa, Lisboa, Portugal, 2Hospital da Luz Learning Health, Luz Saúde, Lisbon, Portugal |
||
Clinical myocardial T1 mapping protocols typically acquire three 2D slices, each obtained in a separate breath-hold; this is only feasible in patients able to do sequential breath-holds. We propose SMS-ProMyoT1 which integrates simultaneous multi-slice (SMS) with open-source ProMyoT1, previously implemented with Pulseq, to obtain all slices in a fast single shot. SMS-ProMyoT1 differs from previous implementations by combining blip-bSSFP with VERSE-multiband for RF modulation, and auto-calibrated blip patterns for a fast self-contained sequence. The method was successfully applied both in a reference phantom and in-vivo. |
| 1341 | Computer 47
|
Feasibility of map recalibration using an in-scan reference system for two MRI mapping cardiac sequences |
| Davide Cicolari1, Domenico Lizio2, Patrizia Pedrotti3, Monica Teresa Moioli2, Alessandro Lascialfari1, Manuel Mariani1, and Alberto Torresin2,4 | ||
1Department of Physics, University of Pavia, Pavia, Italy, 2Department of Medical Physics, ASST Grande Ospedale Metropolitano Niguarda, Milan, Italy, 3Department of Cardiology, ASST Grande Ospedale Metropolitano Niguarda, Milan, Italy, 4Department of Physics, University of Milan, Milan, Italy |
||
In this work we tested the feasibility of MRI relaxation time maps recalibration by employing an in-scan reference system. This ‘belt phantom’ was characterized through NMR standard spectroscopic technique. Scan-dependent recalibrations of the relaxation time maps could be performed relying on the ground-truth NMR values of the phantom, aiming to clinical intra- and inter-center harmonization. The in-scan reference phantom allowed also, together with the analysis of the standard deviation maps (measured as the 68% confidence bound of the fitted relaxation time value), to evaluate the reliability of the maps and the applicability of the recalibration. |
| 1342 | Computer 48
|
Validation of fast in vivo T1 and T2* mapping of tissues in the knee using 3D UTE |
| Maik Rothe1, Andreas Deistung1, Richard Brill1, Walter Alexander Wohlgemuth1, and Alexander Gussew1 | ||
1University clinic and policlinic for Radiology, University Hospital Halle (Saale), Halle (Saale), Germany |
||
In this study we present a 3D high resolution method for fast in vivo T1 and T2* mapping of fast relaxating tissues of the knee. All experiments were performed using a prototypical 3D spoiled gradient echo ultrashort echo time sequence with stack of spirals readout. T1 and T2* values of the mapping technique were validated by reference methods and literature values and showed good agreement for fast T1 and T2* values. The mapping technique enables T1 and T2* quantitation of fast relaxating tissues in a clinically appropriate measurement time of 9 minutes with a submillimeter spatial resolution (0.8mmx0.8mmx0.8mm). |
| 1343 | Computer 49
|
A hybrid EPG-hGLLiM method for accurate fat-signal modeling in skeletal muscle T2 mapping |
| Pierre-Yves Baudin1,2, Ericky Caldas de Almeida Araujo1,2, Harmen Reyngoudt1,2, and Benjamin Marty1,2 | ||
1NMR Laboratory, Institute of Myology, Neuromuscular Investigation Center, Paris, France, 2NMR Laboratory, CEA, DRF, IBFJ, MIRCen, Paris, France |
||
In order to improve the reliability of the muscle T2 as a clinical outcome in clinical studies of neuromuscular disorders, we investigate the accuracy of the fat signal modeling in transverse relaxometry from multi-spin echo imaging. A new approach for muscle T2/fat-fraction mapping is proposed, combining a water signal dictionary with water T2/B1-dependent entries created from EPG simulations, and a hybrid Gaussian Locally Linear Mapping (hGLLiM) to provide B1-dependent fat signals. Preliminary results on a dataset of healthy controls, DMD and IBM patients are consistent with common knowledge, but interesting divergences are found when compared to another recent approach. |
| 1344 | Computer 50
|
Measuring the Impact of Pineapple Juice on Quantitative Liver MRI Metrics cT1, Iron and PDFF at 1.5T and 3T |
| Yi-Chun Wang1,2, Faezeh Sanaei-Nezhad1, Joao Peixoto1, Carolina Fernandes1, Alex Smith1, Matthew Robson1, and Rajarshi Banerjee1 | ||
1Perspectum Ltd., Oxford, United Kingdom, 2University of Oxford, Oxford, United Kingdom |
||
Pineapple juice (PJ) is used in Magnetic Resonance Cholangiopancreatography to suppress gastrointestinal tract signals. To explore how PJ affects multiparametric liver MRI, 30 participants underwent scans before and after ingesting PJ. Image data were analysed with LiverMultiScan (yielding iron corrected T1 (cT1), iron, and PDFF) and statistically compared. The changes post PJ administration were statistically significant for cT1 at both 1.5T and 3T but not for iron and PDFF. However, the repeatability analysis indicates the post PJ administration cT1 changes were smaller than the repeatability limits of agreement, suggesting that PJ has neglect-able clinical effect on multiparametric liver MRI. |
| 1345 | Computer 51
|
Improving annotation accuracy in MRI data using MR Fingerprinting and deep learning |
| Yong Chen1, Rasim Boyacioglu1, Gamage Sugandima Nishadi Weragoda2, Michael Martens2, Chaitra Badve1,3, and Mark Griswold1 | ||
1Radiology, Case Western Reserve University, Cleveland, OH, United States, 2Physics, Case Western Reserve University, Cleveland, OH, United States, 3Radiology, University Hospitals Cleveland Medical Center, Cleveland, OH, United States |
||
In this study, we introduced a new deep learning method to take advantage of both radiologists’ expertise and multi-parametric MR Fingerprinting data to improve annotation accuracy in MRI dataset. A U-Net based convolutional neural network was adopted and each dataset was evaluated multiple times using different combinations of training dataset. Our initial results obtained from a brain tumor dataset demonstrates that the developed method could effectively identify mislabeled tissues and improve annotation accuracy. |
| 1346 | Computer 52
|
Efficient MTsat mapping using sparse MP2RAGE for T1 and M0 measurement with B1+ inhomogeneity correction |
| Christopher D Rowley1,2, Ilana R. Leppert1, Jennifer S.W. Campbell1, Mark C. Nelson1, G Bruce Pike3, and Christine L Tardif1,2,4 | ||
1McConnell Brain Imaging Centre, McGill University, Montreal, QC, Canada, 2Neurology and Neurosurgery, McGill University, Montreal, QC, Canada, 3Hotchkiss Brain Institute and Departments of Radiology and Clinical Neuroscience, University of Calgary, Calgary, AB, Canada, 4Biomedical Engineering, McGill University, Montreal, QC, Canada |
||
In this study we investigate the use of MP2RAGE derived M0 and T1 maps for use in MTsat imaging. These values are compared against maps calculated from the traditional VFA approach. Additionally, the MP2RAGE approach is evaluated with and without sparse sampling and compressed sensing reconstruction. We show that VFA produces elevated T1 values, and consequently lower MTsat compared to MP2RAGE approaches. A model-based ΔB1+ correction was able to remove the dependence of MTsat on ΔB1+. The sparse MP2RAGE provided high quality maps, leading to a full MTsat protocol that was 32% faster than the traditional approach. |
| 1347 | Computer 53
|
Simultaneous mapping of compartment-specific T2 and T2* with Diffusion-PEPTIDE imaging |
| Ting Gong1, Merlin J. Fair2, Kawin Setsompop2,3, and Hui Zhang1 | ||
1Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, 2Radiological Sciences Laboratory, Department of Radiology, Stanford University, Stanford, CA, United States, 3Department of Electrical Engineering, Stanford University, Stanford, CA, United States |
||
Inspired by recent developments in combined relaxometry-diffusion imaging and growing interests in T2* imaging, this study achieves simultaneous compartmental T2 and T2* mapping for the first time. We take advantage of the recently developed diffusion-PEPTIDE imaging technique, which allows for efficient acquisition of diffusion datasets with varying T2/T2* contrast. A relaxometry-diffusion combined microstructure model and a robust multi-stage fitting approach are developed to enable compartmental T2 and T2* mapping. This technique could be a new tool for future neuroimaging studies benefiting from quantification of T2-T2*-diffusion, such as studying iron content related development and pathology. |
| 1348 | Computer 54
|
3D MR-STAT: towards a fast multi-parametric protocol with increased SNR |
| Hongyan Liu1, Oscar van den Heide1, Miha Fuderer1, Cornelis A.T. van den Berg1, and Alessandro Sbrizzi1 | ||
1Computational Imaging Group for MR diagnostics & therapy, Center for Image Sciences, UMC Utrecht, Utrecht, Netherlands |
||
MR-STAT is a framework for simultaneous mapping of quantitative MR parameters from single short scans. In this work, we extend the current 2D Cartesian MR-STAT sequence to 3D acquisitions, in order to achieve higher SNR and isotropic resolution with a scan lasting few minutes. In particular, we implement a prototype 3D MR-STAT sequence for whole brain coverage with isotropic resolution, and test it on gel phantom and in-vivo brain data. Acceleration strategy of the 3D sequence is proposed, and the corresponding retrospective undersampling of the measured in-vivo data is reconstructed. |
| 1349 | Computer 55
|
Simultaneous T1 and T2 mapping with QuantoRAGE: a new MP2RAGE variant using T2-prepared inversion |
| Gabriele Bonanno1,2,3, José P. Marques4, Tobias Kober5,6,7, and Tom Hilbert5,6,7 | ||
1Siemens Healthcare AG, Bern, Switzerland, 2Magnetic Resonance Methodology, Institute of Diagnostic and Interventional Neuroradiology, University of Bern, Bern, Switzerland, 3Translational Imaging Center, sitem-insel, Bern, Switzerland, 4Donders Institute for Brain Cognition and Behaviour, Radboud University, Nijmegen, Netherlands, 5Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, 6Department of Radiology, University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland, 7LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland |
||
T1 and T2 relaxometry provide important quantitative information and can serve as imaging biomarkers thanks to their sensitivity to pathology. However, quantitative imaging requires long scan times for high-resolution whole-brain coverage. Magnetization-prepared approaches combined with fast sequences allow for high isotropic resolution, but they are often biased due to other relaxation mechanisms than the one being probed. To account for this effect, we introduce QuantoRAGE: a new method that uses a T2-prepared inversion within an accelerated MP2RAGE sequence for simultaneous T1 and T2 mapping. Preliminary tests demonstrate the feasibility to obtain high-resolution simultaneous T1 and T2 relaxometry at 3T. |
| 1350 | Computer 56
|
Optimal acquisition settings for simultaneous diffusion kurtosis, free water fraction and T2 estimation |
| Vincenzo Anania1,2, Ben Jeurissen1,3, Jan Morez1, Annemieke E. Buikema1, Thibo Billiet2, Jan Sijbers1,3, and Arnold J. den Dekker1,3 | ||
1imec-Vision Lab, Dept. of Physics, University of Antwerp, Antwerp, Belgium, 2icometrix, Leuven, Belgium, 3µNEURO Research Centre of Excellence, University of Antwerp, Antwerp, Belgium |
||
Fitting the diffusion kurtosis imaging free water elimination model (DKI-FWE) to diffusion MRI data represents an ill-conditioned problem. Fortunately, the conditioning of the model fitting can be improved by explicitly modeling the T2 relaxation dependency of the signal. As a benefit, diffusion and kurtosis metrics robust to partial volume effects can be estimated with conventional techniques. In this work, we use Cramér-Rao lower bound (CRLB) theory to identify optimal acquisition settings that maximize the precision of the model parameter estimates. |
| 1415 | Computer 39
|
Correlating neurite density and synaptic density in the human brain in vivo with diffusion-weighted PET-MR |
| Daan Christiaens1,2, Thomas Vande Casteele2,3,4, Maarten Laroy2,4, Margot Van Cauwenberge2,3,4, Jan Van den Stock3,4, Filip Boeckaert3,4, Stefan Sunaert2,5,6, Mathieu Vandenbulcke3,4, Frederik Maes1,2, and Louise Emsell2,3,4,5 | ||
1Department of Electrical Engineering, ESAT/PSI, KU Leuven, Leuven, Belgium, 2Medical Imaging Research Center, University Hospitals Leuven, Leuven, Belgium, 3Geriatric Psychiatry, UPC KU Leuven, Leuven, Belgium, 4Leuven Brain Institute, Department of Neurosciences, Neuropsychiatry, KU Leuven, Leuven, Belgium, 5Department of Imaging & Pathology, Translational MRI, KU Leuven, Leuven, Belgium, 6Radiology, University Hospitals Leuven, Leuven, Belgium |
||
11C-UCB-J PET offers a unique imaging modality to map synaptic density in the human brain in vivo with high specificity. Here, we investigate its correlation with several diffusion MRI metrics and microstructure model parameters in diffusion-weighted PET-MR. We report moderate negative correlation of 11C-UCB-J uptake with measures of anisotropy, consistent with a hypothesis that higher synaptic density is associated with a more disorganised neurite configuration. We also find weak positive correlation to the intra-axonal signal fraction in cortical grey matter. As such, 11C-UCB-J PET-MR can further the interpretation and in vivo validation of more advanced microstructure models of grey matter. |
| 1416 | Computer 40
|
The impact of learning rate, network size, and training time on unsupervised deep learning for intravoxel incoherent motion (IVIM) model fitting |
| Misha Pieter Thijs Kaandorp1,2, Frank Zijlstra1,2, João P. de Almeida Martins1,2, Christian Federau3,4, and Peter T. While1,2 | ||
1Department of Radiology and Nuclear Medicine, St. Olav’s University Hospital, Trondheim, Norway, 2Department of Circulation and Medical Imaging, NTNU – Norwegian University of Science and Technology, Trondheim, Norway, 3Institute for Biomedical Engineering, University and ETH Zürich, Zurich, Switzerland, 4AI Medical, Zürich, Switzerland |
||
We demonstrate that a high learning rate, small network size, and early stopping in unsupervised deep learning for IVIM model fitting can result in sub-optimal solutions and correlated parameters. In simulations, we show that prolonging training beyond early stopping resolves these correlations and reduces parameter error, providing an alternative to exhaustive hyperparameter optimization. However, extensive training results in increased noise sensitivity, tending towards the behavior of least squares fitting. In in-vivo data from glioma patients, fitting residuals were almost identical between approaches, whereas pseudo-diffusion maps varied considerably, demonstrating the difficulty of fitting D* in these regions. |
| 1417 | Computer 41
|
Improved Modelling of ADC Distributions makes Therapeutic co-clinical Trials possible with very small animal numbers |
| Paul Tar1, N.A. Thacker2, M. Babur2, G. Lipowska-Bhalla2, S. Cheung2, R. Little2, K.J. Williams2, and J.P.B. O'Connor2 | ||
1Cancer, University of Manchester, Manchester, United Kingdom, 2University of Manchester, Manchester, United Kingdom |
||
In oncology, preclinical experiments using MRI often evaluate spatially complex and heterogeneous tumor micro-environments which have non-Gaussian data and small sample sizes, with cohorts typically of 10 animals or less. As a consequence, conventional use of t-tests that evaluate distribution parameters such as means and percentiles can be ineffective. Further, the cohort-level nature of such analyses also limits investigations to groups of tumors rather than identifying individually responding tumors. In contrast, Linear Poisson Modelling (LPM) enables quantitative analysis of complex data, can operate in small data domains and can also provide per-tumor assessments 2.ideal for co-clinical trials. |
| 1418 | Computer 42
|
GRL-MF: The first lego-brick of a cerebellar Mean-Field Model for BOLD signal simulations |
| Roberta Maria Lorenzi1, Alice Geminiani1, Claudia AM Gandini Wheeler-Kingshott1,2,3, Fulvia Palesi1, Claudia Casellato1, and Egidio D'Angelo1,3 | ||
1Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, 2NMR Research Unit, Queen Square Multiple Sclerosis Centre, Department of Neuroinflammation, UCL Queen Square Institute of Neurology, UCL, London, United Kingdom, 3Brain Connectivity Center, IRCCS Mondino Foundation, Pavia, Italy |
||
Following MRI recordings, the connectome can be reconstructed and BOLD-signals simulated extracting relevant parameters on brain organization and dynamics. This requires mathematical models of neural activity in specific brain regions. However, models of the cerebellar circuit are still missing. We present here a biologically-driven mean-field (MF) model summarising the main statistical moments of cerebellar granular layer activity as the first “lego-brick” toward full cerebellar network reconstruction. Once integrated into brain simulators, like Dynamic Causal Modelling and The Virtual Brain, the cerebellar MF models will improve the investigation of neuronal functions at the origin of hemodynamic responses captured by BOLD fMRI. |
| 1419 | Computer 43
|
Evaluation of SS3T-CSD for the analysis of diffusion MRI data - a simulation study |
| Jibin Tang1, Maximilian Pietsch2,3, and Jacques-Donald Tournier2,4 | ||
1Department of Medical Engineering and Physics, School of Biomedical Engineering and Imaging Sciences, Faculty of Life Sciences and Medicine, King's College London, London, United Kingdom, 2Centre for the Developing Brain, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 3MRC Centre for Neurodevelopmental Disorders, King's College London, London, United Kingdom, 4Biomedical Engineering Department, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom |
||
Single-shell three-tissue constrained spherical deconvolution (SS3T-CSD) has been proposed to decompose the dMRI signal into three different tissue types from single-shell data. Here, we evaluate the SS3T-CSD method over a range of different situations using real data and simulations, varying parameters including SNR, partial volume effect, b-value, number of iterations, and tissue response functions. We evaluate their effect on results by comparing with the ground truth and the results estimated from MSMT-CSD. Our results indicate that while SS3T-CSD performs well, it is sensitive to some of these factors and shows some deficiencies in certain situations. |
| 1420 | Computer 44
|
A personalized-NODDI (pNODDI) pipeline increases sensitivity to microstructural alterations in temporal lobe epilepsy |
| Elena Grosso1, Claudia AM Gandini Wheeler-Kingshott1,2,3, Egidio D'Angelo1,3, Paolo Vitali4, and Fulvia Palesi1 | ||
1Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, 2NMR Research Unit, Queen Square Multiple Sclerosis Centre, Department of Neuroinflammation, UCL Queen Square Institute of Neurology, UCL, London, United Kingdom, 3Brain Connectivity Centre Research Unit, IRCCS Mondino Foundation, Pavia, Italy, 4Radiology, IRCCS Policlinico San Donato, San Donato Milanese, Italy |
||
Several mesoscopic multi-compartment diffusion MRI models have been developed to describe the complexity of water molecules diffusion in the brain by making assumptions. One frequently used model is Neurite Orientation Dispersion and Density Imaging (NODDI). Here, the aim is to propose a personalized-NODDI (pNODDI) by making NODDI assumptions subject-specific. Our findings show that NODDI assumptions impact on output metrics depending on brain region and that these metrics are able to discriminate between healthy and temporal lobe epilepsy subjects. pNODDI provides a mean to personalize a successful and clinically feasible model as NODDI, increasing metrics sensitivity to pathological alterations. |
| 1421 | Computer 45
|
Effect of brain tissue deformation on functional MRI signal variations assessed using biomechanical simulations |
| Mahsa Zoraghi1, Nico Scherf2,3, Carsten Jaeger1, Ingolf Sack4, Sebastian Hirsch5,6, Stefan Hetzer5,6, and Nikolaus Weiskopf1,7 | ||
1Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Methods and Development Group Neural Data Science and Statistical Computing, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 3Institute for Medical Informatics and Biometry, Carl Gustav Carus Faculty of Medicine, TU Dresden, Dresden, Germany, 4Department of Radiology, Charité – Universitätsmedizin Berlin, Berlin, Germany, Berlin, Germany, 5Berlin Center for Advanced Neuroimaging, Charité – Universitätsmedizin Berlin, Berlin, Germany, 6Berlin Center for Computational Neuroscience, Berlin, Germany, 7Faculty of Physics and Earth Sciences, Felix Bloch Institute for Solid State Physics, Leipzig University, Leipzig, Germany |
||
Recent studies on the brain suggest a link between brain function and biomechanics of brain. In this study we develop a novel simulation framework to investigate blood flow-induced deformation caused by increased neural activity in brain tissue considering its mechanical properties. We further investigate its impact in simulated fMRI experiments. Our results demonstrate that the displacement in gyrus induced by local volume and stiffness change might not be directly resolved due to limited fMRI resolution, but can lead to artifacts when interpreting measurements at layer-level resolution. Our findings may help to systematically analyze potential resulting artefacts in high resolution fMRI. |
| 1422 | Computer 46
|
Quantitative T1D assessment in lipid membranes: Jeener-Broekaert NMR vs. ihMT MRI |
| Andreea Hertanu1,2, Lucas Soustelle1,2, Ludovic de Rochefort1,2, Axelle Grélard3, Antoine Loquet3, Erick J. Dufourc3, Gopal Varma4, David C. Alsop4, Guillaume Duhamel1,2, and Olivier M. Girard1,2 | ||
1Aix Marseille Univ, CNRS, CRMBM, Marseille, France, 2APHM, Hôpital Universitaire Timone, CEMEREM, Marseille, France, 3CBMN UMR 5248, CNRS, University of Bordeaux, Bordeaux INP, Pessac, France, 4Division of MR Research, Radiology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States |
||
The dipolar order relaxation time (T1D) is a probe of membrane dynamics and microstructure and could serve to further understand the relationship between the myelin membrane integrity and its biological function. In this work, the ability of quantitative inhomogeneous Magnetization Transfer (qihMT) to estimate the several T1D components of a synthetic lipid membrane system, a surrogate for the myelin membrane, was evaluated by comparison with the gold standard method for T1D quantification, the NMR Jeener-Broekaert (JB) sequence. |
| 1423 | Computer 47
|
Susceptibility corrected Proton Resonance Frequency Shift based MR Thermometry of hepatic Microwave Ablation in an in-vivo swine model |
| Marcel Gutberlet1,2, Enrico Pannicke2,3, Inga Bruesch4, Regina Rumpel4, Eva-Maria Wittauer4, Florian W. R. Vondran5, Frank Wacker1,2, and Bennet Hensen1,2 | ||
1Institute of Diagnostic and Interventional Radiology, Medical School Hannover, Hannover, Germany, 2STIMULATE-Solution Centre for Image Guided Local Therapies, Magdeburg, Germany, 3Department Biomedical Magnetic Resonance, Otto-von-Guericke University, Magdeburg, Germany, 4Institute for Laboratory Animal Science and Central Animal Facility, Medical School Hannover, Hannover, Germany, 5Clinic for General, Abdominal and Transplant Surgery, Medical School Hannover, Hannover, Germany |
||
In microwave ablation (MWA), heat induced susceptibility changes impair the assessment of the ablation zones using proton resonance frequency shift based magnetic resonance (MR) thermometry. In this work, these heat related field changes were modelled to improve the accuracy of MR thermometry to monitor microwave ablation. In a study of hepatic MWA in an in-vivo swine model, the proposed method provided increased accuracy to assess the ablation zone compared to uncorrected MR thermometry. |
| 1424 | Computer 48
|
Voxelwise estimation of multiple response functions and fiber orientation distribution functions in diffusion MRI |
| Samuel St-Jean1,2, Alexander Leemans3, Filip Szczepankiewicz1, Christian Beaulieu2, and Markus Nilsson1 | ||
1Clinical Sciences Lund, Lund University, Lund, Sweden, 2University of Alberta, Edmonton, AB, Canada, 3Image Sciences Institute, UMC Utrecht, Utrecht, Netherlands |
||
We present a framework to compute a response function for each voxel using a fingerprinting approach and a per voxel, per compartment estimation of the fiber orientation distribution (FOD) function for diffusion MRI. The method first matches the powder-averaged signal to a database of candidate signals and estimates a unique FOD from this matched signal for each compartment. This is done by turning the candidate signals into continuous Legendre polynomials, leading to a set of linear equations to estimate coefficients of the FODs, without requiring a prior segmentation or averaging multiple response functions. |
| 1425 | Computer 49
|
Novel particle-based spatial stochastic Bloch simulation applied to LDH-mediated pyruvate conversion |
| Dylan Archer Dingwell1,2 and Charles H Cunningham1,2 | ||
1Medical Biophysics, University of Toronto, Toronto, ON, Canada, 2Physical Sciences, Sunnybrook Research Institute, Toronto, ON, Canada |
||
In order to model signal mechanisms relevant to HP 13C MRI of lactate, we developed a novel particle-based MR model which extends the Brownian dynamics simulator Smoldyn. The model performs concurrent calculation of a forward solution of the Bloch equations for simulated particles using quaternion rotations. Reaction kinetics of LDH-mediated conversion of pyruvate to lactate were simulated and compared to a benchtop experiment (LDH activity assay). Modelling of particle compartmentalization and motion were also tested in simulated structures. |
| 1426 | Computer 50
|
Time dependence at ultra-high diffusion weighting reveals fast compartmental exchange in rat cortex in vivo |
| Jonas Lynge Olesen1,2, Andrada Ianus3, Noam Shemesh3, and Sune Nørhøj Jespersen1,2 | ||
1Center of Functionally Integrative Neuroscience (CFIN) and MINDLab, Aarhus University, Aarhus, Denmark, 2Department of Physics and Astronomy, Aarhus University, Aarhus, Denmark, 3Champalimaud Research, Champalimaud Centre for the Unknown, Lisbon, Portugal |
||
The potential effect of compartmental water exchange and somas on the diffusion MRI signal in grey matter tissue are currently open questions with considerable implications for biophysical modelling of grey matter at (pre-)clinical diffusion time scales. Recent studies have begun to tackle these questions and found evidence that compartmental water exchange dominates the diffusion time dependence. Here, we extend the application of the Standard Model with exchange between the neurites and extracellular water (SMEX) to in vivo rat brain and again find evidence for exchange-dominated time dependence with very short neurite residence time on the order of 2 ms. |
| 1427 | Computer 51
|
A novel clinical prostate diffusion MRI approach for combined measurement of ADC and non-monoexponential diffusion |
| Stefan Kuczera1, Fredrik Langkilde1, and Stephan E. Maier1,2 | ||
1Institute of Clinical Sciences, Sahlgrenska Academy, Gothenburg University, Gothenborg, Sweden, 2Department of Radiology, Brigham Women’s Hospital, Harvard Medical School, Boston, MA, United States |
||
A novel clinical protocol for prostate diffusion imaging is presented. The commonly applied strategy of averaging MR signal at a small number of b-values is replaced by the acquisition of a whole range of b-values in conjunction with a model fitting post precessing step. With no increase in acquisition time, our results show that the same image quality can be achieved. As a main benefit, modelling is made more consistent and largely b-value independent for both the simple ADC model, as well as for more complex signal representations such as the kurtosis model. |
| 1428 | Computer 52
|
Understanding Alzheimer's disease through fMRI and deep learning |
| Sarah Wacher1, Jinglei Lv1,2, and Mariano Cabezas1 | ||
1Brain and Mind Centre, The University of Sydney, Sydney, Australia, 2School of Biomedical Engineering, The University of Sydney, Sydney, Australia |
||
Deep learning approaches for Alzheimer's disease (AD) classification have primarily focused on modelling the structural changes associated with the condition, neglecting the changes in functional brain dynamics. These functional changes may be detectable earlier than structural atrophy providing an avenue for earlier diagnosis and treatment. We therefore proposed a convolutional neural network combined with a long short-term memory unit to decode fMRI signals. The model was able to classify AD from healthy control with a balanced accuracy of 0.69. Whilst there is room to improve network performance, the study already provides promising insights into the possibilities of resting-state fMRI classification. |
| 1429 | Computer 53
|
Classification of Chemotherapy-Related Subjective Cognitive Complaints in Breast Cancer Using Multi-Level Features of Functional MRI |
| Lei Wang1 and Fuqing Zhou1 | ||
1Department of Radiology, The First Affiliated Hospital of Nanchang University, Nanchang, China |
||
Chemotherapy-related cognitive impairment (CRCI), especially subjective cognitive complaints (SCC), had been often reported in breast cancer survivors, which affects their life and work. The identification of biomarkers for early diagnosis and prognosis prediction of SCC remains a crucial challenge of important clinical implications. The resting-state functional magnetic resonance imaging (rs-fMRI) has been widely used to detect abnormalities of brain activity in CRCI. The machine learning method combined with rs-fMRI features could effectively identify breast cancer survivors with chemotherapy-related SCC from healthy controls (HC). |
| 1430 | Computer 54
|
Multi-Channel Deep Learning for IDH Mutation Status Prediction in Gliomas：A Multimodal Approach |
| Jin Wang1, Jiayang Deng2, Rui Wang1, and Jing Zhang1 | ||
1Department of Magnetic Resonance, LanZhou University Second Hospital, Lanzhou, China, 2Vipshop (China) Co. LTD, Guangzhou, China |
||
Isocitrate dehydrogenase (IDH) mutation status has emerged as an important prognostic marker in gliomas.Currently, reliable IDH mutation determination requires invasive surgical procedures. Various studies have demonstrated the efficacy of deep learning in classify IDH status using Magnetic resonance images（MRI）data.In this study we propose a multi-channel architecture of 3D convolutional neural networks (CNNs) based on deep learning to predict the IDH status.We utilize traditional structures imaging and various diffusivity metric maps derived from diffusion tensor imaging (DTI) as input to the network.The final model achieved the AUC value of 0.93. |
| 1431 | Computer 55
|
MR-only staging of osteonecrosis of the femoral head (ONFH) using self-supervised learning for multiple MR protocols |
| Bomin Kim1, Geun Young Lee2, and Sung-Hong Park1 | ||
1Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, 2Deparment of Radiology, Chung-Ang University Hospital, Seoul, Korea, Republic of |
||
Diagnosing stages of osteonecrosis of the femoral head (ONFH) based on MR images can reduce additional cost and radiation exposure caused by CT scan. We propose a deep learning network that enables 4-way classification of ONFH stages, utilizing the information from MR images with different contrasts and planes. Given the limited number of available data, we enhanced the network performance by using self-supervised learning based on MR-to-CT translation task, which increased AUC significantly. We also investigated the diagnostic results from different MR protocols, and obtained more precise and robust results by combining them. |
| 1432 | Computer 56
|
Towards a personalized MRgFUS treatment for tremor disorders: A study on the number of ablations using deep learning and structural connectivity |
| Zihao Tang1,2, Mariano Cabezas2, Kain Kyle2,3, Arkiev D'Souza2, Stephen Tisch4, Ben Jonker4, Yael Barnett4, Joel Maamary4, Jerome Maller5, Michael Barnett2,3, Weidong Cai1, and Chenyu Wang2,3 | ||
1School of Computer Science, University of Sydney, Sydney, Australia, 2Brain and Mind Centre, University of Sydney, Sydney, Australia, 3Sydney Neuroimaging Analysis Centre, Sydney, Australia, 4St Vincent's Hospital Sydney, Sydney, Australia, 5GE Healthcare Australia, Melbourne, Australia |
||
Disabling tremor is the most common symptom of tremor-dominated Parkinson’s disease (PD) patients. Recently, MR guided focused ultrasound (MRgFUS) has been applied in the clinical environment to treat tremor. However, patients can have different treatment responses to the ablation. Tremor suppression can be observed on some patients after ablating the ventral intermediate nucleus (Vim), while others require further ablations. To provide a tailored treatment, a deep learning technique is introduced in this work to predict whether a subject undergoing MRgFUS will respond to a single ablation in the VIM or require additional lesioning in other regions. |
| 1433 | Computer 57
|
Improved assessment of fetal ocular pathologies in MRI using ocular ratios and machine-learning multi-parametric classification |
| Netanell Avisdris1,2, Daphna Link Sourani1, Liat Ben Sira3,4,5, Leo Joskowicz2, Gustavo Malinger4,6, Simcha Yagel7, Elka Miller8, and Dafna Ben Bashat1,4,5 | ||
1Sagol Brain Institute, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, 2School of Computer Science and Engineering, The Hebrew University of Jerusalem, Jerusalem, Israel, 3Division of Pediatric Radiology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, 4Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, 5Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel, 6Division of Ultrasound in Obstetrics & Gynecology, Lis Maternity Hospital, Tel Aviv, Israel, 7Obstetrics and Gynecology Division, Hadassah Hebrew University Medical Center, Jerusalem, Israel, 8Medical Imaging, CHEO, University of Ottawa, Ottawa, ON, Canada |
||
Brain MRI of 296 fetuses (22-40 weeks’ gestational-age, normal n=244, hypo/hyper-telorism n=52 were included. Binocular, interocular, and ocular diameters, and ocular volume were measured using automatic methods. Two new parameters, binocular-ratio and interocular-ratio, were defined. In normal fetuses, all four measurements increased with gestational-age. However, constant values were detected across all gestational-ages of binocular-ratio=1.56±0.05 and interocular-ratio=0.60±0.05. A random-forest classifier achieved the best results (out of eight classifiers) with AUC-ROC=0.90±0.03 for classification between normal and fetuses with hypo/hyper-telorism. mainly based on the two new ratios. Machine-learning multi-parametric classification and the new ratios are suggested to be used in routine clinical practice. |
| 1434 | Computer 58
|
Radiomics signature from multiparametric MRI as early in-vivo biomarkers for pseudoprogression in recurrent glioblastoma patients. |
| Lucie Piram1, Acquitter Clément2, Julia Gilhodes3, Umberto Sabatini4, Elizabeth Cohen-Jonathan Moyal1,5, Benjamin Lemasson2, and Soleakhena Ken5,6,7 | ||
1Department of Radiotherapy, Institut Universitaire du Cancer de Toulouse - Oncopole, Toulouse, France, 2Grenoble Institut des Neurosciences, Grenoble, France, 3Department of Clinical Trials, Institut Universitaire du Cancer de Toulouse - Oncopole, Toulouse, France, 4Dipartimento di Scienze Mediche e Chirurgiche, Università Magna Graecia, Catanzaro, Italy, 5U1037, RADOPT Team, Cancer Research Center of Toulouse, Toulouse, France, 6Department of Engineering and Medical Physics, Institut Universitaire du Cancer de Toulouse - Oncopole, Toulouse, France, 7MINDS Team UMR 5505, Institut de Recherche en Informatique de Toulouse, Toulouse, France |
||
Radiomic features computed from multiparametric MRI were found to be relevant as early in-vivo biomarkers for pseudoprogression evaluation in recurrent glioblastoma patients. At baseline, predictive biomarker for pseudoprogression outcome was related to kurtosis parameter of FLAIR histogram plotted from abnormal hyper-intense signal area. When considering variation between baseline and first event (either pseudoprogression or true progression), four early biomarkers were found for entropy of T1-weighted, T1-weighted-post-contrast morphological MRI and Apparent Diffusion Coefficient maps derived from diffusion-weighted MRI. Such early in-vivo biomarkers easily computed from automatic segmentation and first order radiomics analysis could be useful for the assessment of treatment response. |
| 1435 | Computer 59
|
MRI based Radiomics for Distinguishing IDH-mutant from IDH wild-type Grade-4 Astrocytomas |
| Seyyed Ali Hosseini1, Isaac Shiri2, Ghasem Hajianfar3, Stephen Bagley4, MacLean Nasrallah5, Donald M O’Rourke 6, Suyash Mohan7, and SANJEEV CHAWLA7 | ||
1Medical physics and biomedical engineering, Tehran university of medical science, Karaj, Iran (Islamic Republic of), 2Medical Imaging, University of Geneva, Geneva, Switzerland, 3Cardiovascular Medical and Research Center, Iran University of Medical Science, Tehran, Iran (Islamic Republic of), 4Hematology-Oncology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, United States, 5Pathology and Lab Medicine, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, United States, 6Neurosurgery, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, United States, 7Radiology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, United States |
||
To distinguish IDH-mutant from IDH wild-type grade-4 astrocytomas, conventional MR imaging (post-contrast T1-weighted and T2-Flair images) was acquired from 57 patients [IDH-mutant (n=23) and IDH-wild-type (n=34) grade-4 astrocytomas]. Post-contrast T1-weighted and T2-FLAIR images were resliced, resampled and co-registered. Neoplasms were segmented into whole tumor, enhancing region, central necrotic region, edema region, and core tumor (contrast enhancing + necrotic region). A total of 105 first-, second-, higher-order and shape based radiomics features were extracted from each ROI. Readily interpretable and quantitative features from different sub-regions of neoplasms were observed with high diagnostic performances in distinguishing IDH-mutant from IDH wild-type grade-4 astrocytomas. |
| 1436 | Computer 60
|
Exploring Hybrid CNN-Transformer for Schizophrenia Classification using Structural MRI |
| Vishwanatha Mitnala Rao1, Junhao Zhang1, and Jia Guo2 | ||
1Department of Biomedical Engineering, Columbia University, New York, NY, United States, 2Department of Psychiatry, Columbia University, New York, NY, United States |
||
Schizophrenia diagnosis is clinically difficult due to the lack of biomarkers associated with the disease. While machine learning algorithms and convolutional neural networks (CNNs) have found success using neuroimaging inputs to diagnose the disease, they have historically not performed or generalized well enough for clinical use. We propose 3D-MIC-Transformer, the first transformer-based deep learning architecture applied to neurological disease classification that demonstrates state-of-the-art schizophrenia classification performance and generalization using structural MRI inputs. 3D-MIC-Transformer outperforms prior CNN implementations (AUROC: 0.985, accuracy: 0.933), and we believe 3D-MIC-Transformer can serve as a backbone for other disease classification tasks in the future. |
| 1437 | Computer 61
|
Differentiation of NF2 Loss and S100 Positivity in Meningioma Using Dynamic Susceptibility Contrast MRI with Machine Learning at 3T |
| Buse Buz-Yalug1, Ayca Ersen Danyeli2,3, Kubra Tan4, Ozge Can5, Necmettin Pamir3,6, Alp Dincer3,7, Koray Ozduman3,8, and Esin Ozturk-Isik1 | ||
1Institute of Biomedical Engineering, Bogazici University, Istanbul, Turkey, 2Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 3Center for Neuroradiological Applications and Reseach, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 4Health Institutes of Turkey, Istanbul, Turkey, 5Department of Medical Engineering, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 6Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 7Department of Radiology, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 8Department of Neurosurgery, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey |
||
Meningiomas are the most frequent primary intracranial and spinal tumors. In this work, we studied the relative cerebral blood volume (rCBV) changes in the meningiomas with NF2 loss and S100 positivity. Meningiomas with NF2-L had lower rCBV than NF2-NL, and S100 positive group had also lower rCBV than S100 negative group. The highest classification accuracies obtained using machine learning applications were 75.6% for NF2 molecular subsets and 75.2% for S100 molecular subsets. |
| 1438 | Computer 62
|
Deep Learning based Total Kidney Volume Segmentation in Autosomal Dominant Polycystic Kidney Disease |
| Anish Raj1, Fabian Tollens2, Anika Strittmatter1, Laura Hansen1, Dominik Noerenberg2, and Frank G Zöllner1 | ||
1Computer Assisted Clinical Medicine, Mannheim Institute for Intelligent Systems in Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, 2Department of Clinical Radiology and Nuclear Medicine, Medical University Center Mannheim, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany |
||
The total kidney volume (TKV) increases with ADPKD progression and hence, can be used to quantify disease progression. The TKV calculation requires accurate delineation of kidney volumes which is usually performed manually by an expert physician. However, this is time consuming and automated segmentation is warranted, e.g., using deep learning. The implementation of the latter is usually hindered due to a lack of large, annotated datasets.In this work, we address this problem by implementing the cosine loss function and a technique called Sharpness Aware Minimization (SAM) into the U-Net to improve TKV estimation in small sized datasets. |
| 1497 | Computer 47
|
Two-dimensional population receptive field mapping of human primary somatosensory cortex |
| Michael Asghar1, Rosa Sanchez Panchuelo2, Denis Schluppeck1, and Susan Francis1 | ||
1University of Nottingham, Nottingham, United Kingdom, 2University Hospitals Birmingham NHS Foundation Trust, Birmingham, United Kingdom |
||
High-resolution fMRI data of somatosensory cortex were obtained at 7T (n=7 participants) using a 4x4 grid of piezoelectric stimulators. Population receptive field (pRF) mapping was performed using different model specifications (1D Gaussian within-digit, 1D Gaussian between-digit, 2D Gaussian of within-digit and between-digit, Unconstrained model) and pRF sizes were estimated. Grouping pRF response by 4 preferred locations within digits (along tip-base axis) or by Brodmann areas revealed more elliptical pRFs within digit representations of D2 to D4 and more circularly symmetric pRFs for D5. We also found relatively larger pRF sizes in D5 and towards the base of the digits. |
| 1498 | Computer 48
|
GLCM texture analysis of knee cartilage T2 maps: machine learning based selection of important features |
| Veronika Janacova1, Pavol Szomolanyi1, Dominik Vilimek1,2, Siegfried Trattnig1,3,4, and Vladimir Juras1 | ||
1High Field MR Centre, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria, 2Department of Cybernetics and Biomedical Engineering, VSB–Technical University of Ostrava, Ostrava, Czech Republic, 3CD laboratory for Clinical Molecular MR imaging (MOLIMA), Vienna, Austria, 4Institute for Clinical Molecular MRI in the Musculoskeletal System, Karl Landsteiner Society, Vienna, Austria |
||
Texture analysis of quantitative T2 maps in combination with machine learning was explored as a tool for classification and prediction of various conditions in musculoskeletal (MSK) research. We explored random forest classification algorithm as a tool for identification of important texture features for the classification of MACT grafts and native cartilage twelve months after surgery. Our model performed with high accuracy (84.6%) and identified features with highest importance were: cluster prominence, sum average, autocorrelation and correlation. |
| 1499 | Computer 49
|
Application of Two Denoising Methods to the Diffusion-Weighted 129Xe MRI and Dynamic Ventilation 19F/129Xe MRI Data Obtained from Normal Rats |
| Elise Noelle Woodward1, Gregory Lemberskiy2, Matthew S Fox1,3, and Alexei Ouriadov1,3 | ||
1Physics and Astronomy, Western University, London, ON, Canada, 2New York University Grossman School of Medicine, New York, NY, United States, 3Lawson Health Research Institute, London, ON, Canada |
||
In this study, we demonstrate the effectiveness of two de-noising methods on diffusion-weighted and dynamic lung imaging in healthy rats using 129Xe and 19F. While the results showed a insignificant difference for denoising method 1 (MP-PCA), de-noising method 2 (MP-PCA preceded by hard-thresholding over a nuclear norm in k-space) showed significant difference in the overall mean Lm and LmD values for ventilation and r (fractional ventilation estimate) for ventilation images. |
| 1500 | Computer 50
|
Mapping Free Water in the Brain Using Multiparameter qMRI Combined with an Iterative Fast Exchange Relaxation Model |
| Ryan McNaughton1,2, Ning Hua2, Lei Zhang3, David Kennedy4, Karl Kuban2, and Hernan Jara2 | ||
1Mechanical Engineering, Boston University, Boston, MA, United States, 2Boston University Medical Center, Boston, MA, United States, 3University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 4University of Massachusetts, Worcester, MA, United States |
||
Purpose: To develop an integrated theoretical model and computer algorithms for mapping individually and simultaneously the spatial distributions of free water, alongside myelin and iron. Theory: We solve a fast exchange relaxation model effectively decomposing qMRI maps of nPD, R1, and R2 into maps of free water, myelin, and iron. Results: The image processing technique generates maps of free water, myelin, and iron in the brains of adolescents born extremely preterm. Conclusion: A theoretical framework and image processing pipeline for mapping the distribution of free water has been developed and tested with a cohort of adolescents born extremely preterm. |
| 1501 | Computer 51
|
MRI Monitoring of Changes in Kidney Size during Pathophysiological Intervention |
| Thomas Gladytz1, Ehsan Tasbihi1, Jason M. Millward1, Kathleen Cantow2, Luis Hummel2, Joao S. Periquito1,2, Sonia Waiczies1, Erdmann Seeliger2, and Thoralf Niendorf1,3 | ||
1Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbrueck Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany, 2Institute of Physiology, Charité - Universitätsmedizin Berlin, Berlin, Germany, 3Experimental and Clinical Research Center, a joint cooperation between the Charité - Universitätsmedizin Berlin and the Max Delbrueck Center for Molecular Medicine, Berlin, Germany |
||
Several renal pathologies are associated with changes in kidney size, offering an opportunity for magnetic resonance imaging (MRI) biomarkers of disease. An automated bean-shaped model was developed for kidney size measurements in rats using parametric MRI (T2, T2* mapping). The ABSM approach was applied to longitudinal renal size quantification during pathophysiologically relevant interventions affecting kidney size. A precision and accuracy similar to manual segmentation was achieved allowing size changes of 2% to be detected reliably. This can potentially be instrumental for developing MRI-based diagnostic tools for various kidney disorders and for gaining new insight into mechanisms of renal pathophysiology. |
| 1502 | Computer 52
|
Characterizing exchange dynamics in a urea phantom via T2-T2 relaxometry using a non-linear least squares fitting method |
| Ericky Caldas de A Araujo1,2, Benjamin Marty1,2, and Harmen Reyngoudt1,2 | ||
1Neuromuscular investigation center, Institute of Myology, Paris, France, 2NMR Laboratory, CEA/DRF/IBFJ/MIRCen, Paris, France |
||
The multiexponential T2 of water in biological tissue is known to reflect microscopic anatomical compartmentation. T2-T2 correlation relaxometry allows characterizing compartmental sizes, intrinsic T2 values and exchange rates, which are of upmost clinical relevance. However, inversion of relaxation data into T2 spectra is an ill-posed problem. Regularized Inverse Laplace transform (rILT) provides stable solutions, but these are penalized by low spectral resolution and relatively high computational complexity. Here we do T2-T2 relaxometry of a urea solution and show that, for such bi-compartment system, non-linear least squares fitting provides solutions that are more accurate while avoiding the difficulties related to rILT. |
| 1503 | Computer 53
|
Estimation of numerical substrate properties with compartmentalized models from Monte-Carlo simulated DW-MRI signals |
| Rémy Gardier1, Juan Luis Villarreal Haro1, Erick Jorge Canales-Rodriguez1, Gabriel Girard1,2,3, Jonathan Rafael-Patiño1,3, and Jean-Philippe Thiran1,2,3 | ||
1Signal Processing Laboratory (LTS5), Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 2CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 3Radiology Department, Centre Hospitalier Universitaire Vaudois and University of Lausanne, Lausanne, Switzerland |
||
For over a decade, microstructure imaging has been a hot research topic in DW-MRI. Tissue complexity compelled researchers to make assumptions about certain properties. The inverse problem of microstructure imaging, in particular, is ill-posed, and current methods fix some parameters to reduce the solution space. In this abstract, we look at how intracellular and extracellular diffusion coefficients affect the estimation of compartmentalized model parameters. We show that robust estimation of some parameters does not extend to all parameters using Monte-Carlo simulations in impermeable substrates with multiple diffusivities, and we identified the extracellular compartment as the most influential on estimation quality. |
| 1504 | Computer 54
|
Isometric Contractions of the Quadriceps muscle: Strain and Strain Tensor Mapping using Velocity Encoded Phase Contrast Imaging. |
| Shantanu Sinha1, Ryan Hernandez2, Vadim Malis1, T Oda3, and Usha Sinha2 | ||
1Radiology, UC San Diego, San Diego, CA, United States, 2Physics, San Diego State University, San Diego, CA, United States, 3Hyogo University of Teacher Education, Kato, Hyogo, Japan |
||
The quadriceps muscle plays an extremely critical role in various aspects of human locomotion. Changes in its function from various disease such as aging, muscular disuse or modified gait can degrade mobility. VE-PC imaging provides a convenient way to study muscle motion and assess the relative contribution of the thigh muscles to force production. This study is focused on deriving the strain and strain rate tensors to explore patterns along the fiber direction and cross-section. We were able to successfully obtain strain tensor images as well as to follow the temporal variation of these indices in different muscle compartments. |
| 1505 | Computer 55
|
A Hybrid IVIM-DTI Model with Optimized Acquisition Time |
| Sam Sharifzadeh Javidi1 and Hamidreza Salighehrad2,3 | ||
1Department of Medical Physics and Biomedical Engineering, Tehran University of Medical Sciences, Tehran, Iran (Islamic Republic of), 2Department of Medical Physics and Biomedical Engineering, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran (Islamic Republic of), 3Quantitative MR Imaging and Spectroscopy Group, Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences, Tehran, Iran (Islamic Republic of) |
||
IVIM-DTI imaging is capable of revealing both structural and functional maps. We conduct this study to address the long acquisition time of IVIM-DTI imaging and the inaccuracy of estimates of the model’s outputs. Diffusion-weighted images were acquired at 11 b-values and 64 orientations. We used the Kalman filter to reach higher accuracy in a lower number of images. The resulting maps indicated that diffusion maps and pseudodiffusion maps for a healthy case are of highly visual similarity. Our results also confirmed that achieving a good accuracy is possible with just half number of images. |
| 1506 | Computer 56
|
Temporally improved volume registration by data driven subvolume interpolation of simultaneous multislice EPI time series |
| Oscar Albert Dabrowski1, Oscar Albert Dabrowski1, Sébastien Courvoisier2, Jean-Luc Falcone3, Antoine Klauser2, Julien Songeon3, Michel Kocher2, Bastien Chopard3, and Francois Lazeyras2 | ||
1Computer science, University of Geneva, Geneva, Switzerland, 2CIBM, Geneva, Switzerland, 3University of Geneva, Geneva, Switzerland |
||
In the context of EPI time series acquisition, fast motion occurring below the Nyquist limit (< 1 TR), present a problem for accurate registration by 3D rigid-body algorithms. A method for temporally upsampling SMS EPI acquisitions is proposed. EPI volumes are splitted into subvolumes according to the temporal acquisition order of SMS and interpolated using a data driven approach based on pairwise affine transforms and linear interpolation. A proof of concept based EPI acquisitions of controlled head choreographies shows that fast motion below the Nyquist rate can be more accurately estimated by our method. |
| 1507 | Computer 57
|
Understanding T1 in heterogeneous systems: Extending the two-pool model to fractional order |
| Luke Reynolds1, Alex MacKay1,2,3, and Carl Michal1 | ||
1Physics & Astronomy, University of British Columbia, Vancouver, BC, Canada, 2Radiology, University of British Columbia, Vancouver, BC, Canada, 3UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada |
||
Reliably quantifying longitudinal relaxation in heterogenous systems like white matter is challenging because measured parameters are sensitive to details of the measurement technique. In particular, cross-relaxation/magnetization exchange between aqueous and non-aqueous tissue components may lead to multi-exponential relaxation during inversion recovery, depending on the difference in the pools’ initial magnetizations. We use a generalization of the Bloch equations to fractional order to show that the additional component stemming from this exchange is better described by a stretched Mittag-Leffler function than a standard exponential in heterogenous systems. This approach may provide additional information about the material structure. |
| 1508 | Computer 58
|
How modeling lactate diffusion may inform on its cellular compartmentation? An initial study |
| Sophie Malaquin1 and Julien Valette1 | ||
1Université Paris-Saclay, Commissariat à l'Energie Atomique et aux Energies Alternatives (CEA), Centre National de la Recherche Scientifique (CNRS), Molecular Imaging Research Center (MIRCen), Laboratoire des Maladies Neurodégénératives, Fontenay aux Roses, France |
||
Brain lactate compartmentation in neurons, astrocytes and the extracellular space is presumably of critical importance, but remains impossible to assess non-invasively. DW-MRS of lactate might capture lactate fractions in each compartment, provided these compartments induce different diffusion properties. Here we propose a framework where diffusion of cell-specific metabolites (Ins and NAA) is modeled to predict lactate diffusion in astrocytes and neurons. We use some “realistic” (spheres+cylinders) and simplified (cylinders) models of cell microstructure to show that lactate fractions in each compartment can be accurately estimated, despite inaccurate microstructure modeling. Models are finally tested on in vivo mouse brain data. |
| 1509 | Computer 59
|
Characterization of relationship between MR microstructural parameters and memory in older subjects |
| Scott Peltier1,2, Michelle Karker1, Jon-Fredrik Nielsen1,2, Navid Seraji-Bozorgzad3, Henry Paulson3, Bruno Giordani3,4, and Benjamin M. Hampstead4,5 | ||
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, 2Functional MRI Laboratory, University of Michigan, Ann Arbor, MI, United States, 3Neurology, University of Michigan, Ann Arbor, MI, United States, 4Psychiatry, University of Michigan, Ann Arbor, MI, United States, 5Mental Health Service, VA Ann Arbor Healthcare System, Ann Arbor, MI, United States |
||
This work examined the relationship between white matter microstructure parameters calculated by NODDI and a composite measure of memory function. Connectome predictive modeling was done on fractional anisotropy, orientation dispersion, and neurite density maps. Significant relationships were found for all parameters, with orientation dispersion having the highest correlation between predicted and observed memory scores. |
| 1510 | Computer 60
|
All you need are DICOM images |
| Guanxiong Luo1, Moritz Blumenthal1, Xiaoqing Wang1,2, and Martin Uecker1,2,3,4 | ||
1University Medical Center Göttingen, Göttingen, Germany, 2German Centre for Cardiovascular Research (DZHK), Partner Site Göttingen, Göttingen, Germany, 3Institute of Medical Imaging, Graz University of Technology, Graz, Austria, 4Cluster of Excellence "Multiscale Bioimaging: from Molecular Machines to Networks of Excitable Cells'' (MBExC), University of Göttingen, Göttingen, Germany |
||
Most deep-learning-based reconstructions methods need predefined sampling patterns and precomputed coil sensitivities for supervised training, limiting their later use in applications under different conditions. Furthermore, only the magnitude images are always stored in DICOM format in the Picture Archiving and Communication System (PACS) of a typical radiology department. That means that raw k-space data is usually not available. This work focuses on how to extract prior knowledge from magnitude images (DICOM) and how to apply the extracted prior to reconstruct images from k-space multi-channel data sampled with different schemes. |
| 1511 | Computer 61
|
Brain age prediction using fusion deep learning combining pre-engineered and convolution-derived features |
| HeeJoo Lim1,2, Eunji Ha3, Suji Lee3, Sujung Yoon3,4, In Kyoon Lyoo3,4,5, and Taehoon Shin1,2 | ||
1Division of Mechanical and Biomedical Engineering, Ewha Womans University, Seoul, Korea, Republic of, 2Graduate Program in Smart Factory, Ewha Womans University, Seoul, Korea, Republic of, 3Ewha Brain Institute, Ewha Womans University, Seoul, Korea, Republic of, 4Department of Brain and Cognitive Sciences, Ewha Womans University, Seoul, Korea, Republic of, 5Graduate School of Pharmaceutical Sciences, Ewha Womans University, Seoul, Korea, Republic of |
||
Prediction of biological brain age is important as its deviation from chronological age can serve as a biomarker for degenerative neurological disorders. In this study, we suggest novel fusion deep learning algorithms which combine pre-engineered features and convolutional neural net (CNN) extracted features of T1-weighted MR images. Over all backbone CNN architectures, fusion models improved prediction accuracy (mean absolute error (MAE) = 3.40–3.52) compared with feature-engineered regression (MAE = 4.58–5.15) and image-based CNN (MAE = 3.60–3.95) alone. These results indicate that using both features derived from convolution and pre-engineering can complement each other in predicting brain age. |
| 1512 | Computer 62
|
Microbleed detection in autopsied brains from community-based older adults using microbleed synthesis and deep learning |
| Grant Nikseresht1, Ashish A. Tamhane2, Carles Javierre-Petit3, Arnold M. Evia2, David A. Bennett2, Julie A. Schneider2, Gady Agam1, and Konstantinos Arfanakis2,3 | ||
1Department of Computer Science, Illinois Institute of Technology, Chicago, IL, United States, 2Rush Alzheimer's Disease Center, Rush University Medical Center, Chicago, IL, United States, 3Department of Biomedical Engineering, Illinois Institute of Technology, Chicago, IL, United States |
||
Automated cerebral microbleed (CMB) detection on ex-vivo MRI is key to enabling MRI-pathology studies in large community-based cohorts where manual CMB annotation is time consuming and prone to error. The aim of this study is to develop a CMB detection algorithm to aid in the quantization and localization of CMBs on ex-vivo T2*-weighted gradient echo MRI in community-based cohorts. A CMB synthesis algorithm is proposed and synthetic CMBs are used to train a neural network for CMB detection. A model trained with both synthetic and real data is shown to outperform models trained on synthetic or real data alone. |
| 1513 | Computer 63
|
Synthetic T2-weighted fat sat delivers valuable information on spine pathologies: multicenter validation of a Generative Adversarial Network |
| Sarah Schlaeger1, Katharina Drummer1, Malek El Husseini1, Florian Kofler1,2,3, Nico Sollmann1,4,5, Severin Schramm1, Claus Zimmer1, Dimitrios C. Karampinos6, Benedikt Wiestler1, and Jan S. Kirschke1 | ||
1Department of Diagnostic and Interventional Neuroradiology, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany, 2Department of Informatics, Technical University of Munich, Munich, Germany, 3TranslaTUM - Central Insitute for Translational Cancer Research, Technical University of Munich, Munich, Germany, 4TUM-NeuroImaging Center, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany, 5Department of Diagnostic and Interventional Radiology, University Hospital Ulm, Ulm, Germany, 6Department of Diagnostic and Interventional Radiology, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany |
||
Generative Adversarial Networks (GANs) can synthesize missing Magnetic Resonance (MR) contrasts from existing MR data. In spine imaging, sagittal T2-w fat sat (fs) sequences are an important additional MR contrast next to conventional T1-w and T2-w sequences. In this study, the diagnostic performance of a GAN-based, synthetic T2-w fs is evaluated in a multicenter dataset. By comparing the synthetic T2-w fs with its true counterpart regarding ability to detect spinal pathologies not seen on T1-w and non-fs T2-w, diagnostics agreement, and image and fs quality our work shows that a synthetic T2-w fs delivers valuable information on spine pathologies. |
| 1514 | Computer 64
|
Prostate Cancer Risk Assessment using Fully Automatic Deep Learning in MRI: Integration with Clinical Data using Logistic Regression Models |
| Adrian Schrader1,2, Nils Bastian Netzer1,2, Magdalena Görtz3, Constantin Schwab4, Markus Hohenfellner3, Heinz-Peter Schlemmer1, and David Bonekamp1 | ||
1Division of Radiology, German Cancer Research Center, Heidelberg, Germany, 2Heidelberg University Medical School, Heidelberg, Germany, 3Department of Urology, University of Heidelberg Medical Center, Heidelberg, Germany, 4Department of Pathology, University of Heidelberg Medical Center, Heidelberg, Germany |
||
For patients with clinical suspicion for significant prostate cancer, the decision to undergo prostate biopsy can be supported by calculating the individual risk profile using demographic and clinical information along with multiparametric MRI assessment. We could show that the prediction performance of an established risk calculator remained stable after substituting manual PI-RADS scores for assessments from a fully automated deep learning system. Combining deep learning and PI-RADS resulted in significant improvements over using only PI-RADS. Complementary information that deep learning models are able to extract enable synergies with radiologists to improve individual risk predictions. |
| 1515 | Computer 65
|
Multi-parametric versus bi-parametric prostate MRI for deep learning: Marginal benefits from adding dynamic contrast-enhanced images |
| Nils Bastian Netzer1,2, Adrian Schrader1,2, Magdalena Görtz3, Constantin Schwab4, Markus Hohenfellner3, Heinz-Peter Schlemmer1, and David Bonekamp1 | ||
1Radiology, German Cancer Research Center, Heidelberg, Germany, 2Heidelberg University Medical School, Heidelberg, Germany, 3Urology, University of Heidelberg Medical Center, Heidelberg, Germany, 4Pathology, University of Heidelberg Medical Center, Heidelberg, Germany |
||
The value of dynamic contrast enhanced MRI (DCE) for the diagnosis of prostate cancer is unclear and has not yet been investigated in the context of deep learning. We trained 3D U-Nets to segment prostate cancer on bi-parametric MRI and on DCE images of 761 exams. On a test set of 191 exams, the bi-parametric baseline achieved a ROC AUC of 0.89, showing a higher specificity that clinical PI-RADS at a sensitivity of 0.9. Additional improvement could be achieved by fusing bpMRI and DCE predictions, resulting in a ROC AUC of 0.9. |
| 1516 | Computer 66
|
Differentiation of IDH and TERTp mutations in Glioma Using Dynamic Susceptibility Contrast MRI with Machine Learning at 3T |
| Buse Buz-Yalug1, Ayca Ersen Danyeli2,3, Cengiz Yakicier3,4, Necmettin Pamir3,5, Alp Dincer3,6, Koray Ozduman3,7, and Esin Ozturk-Isik1 | ||
1Institute of Biomedical Engineering, Bogazici University, Istanbul, Turkey, 2Department of Medical Pathology, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 3Center for Neuroradiological Applications and Reseach, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 4Department of Molecular Biology and Genetics, Istanbul, Turkey, 5Department of Neurosurgery, Acibadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 6Department of Radiology, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey, 7Department of Neurosurgery, Acıbadem Mehmet Ali Aydinlar University, Istanbul, Turkey |
||
Isocitrate dehydrogenase (IDH) and telomerase reverse transcriptase promoter (TERTp) mutations highly affect the clinical outcome in gliomas. The aim of this study was to identify IDH and TERTp mutations in glioma patients using machine learning approaches on relative cerebral blood volume (rCBV) maps obtained from dynamic susceptibility contrast MRI (DSC-MRI). The highest classification accuracy was 87.2% (sensitivity = 85.7%, specificity = 88.9%) for the IDH subgroup, 81.8% accuracy (sensitivity = 77.5%, specificity = 86.4%) was obtained for classifying the TERTp subgroup. Additionally, a classification accuracy of 89.6% (sensitivity = 88.3%, specificity = 91.2%) was obtained for identifying the TERTp-only gliomas. |
| 1517 | Computer 67
|
Machine learning based classification of major depressive disorder using clinical symptom scales and ultrahigh field MRI features |
| Gaurav Verma1, Xin Xing2, Yael Jacob3, Bradley N Delman4, James Murrough3, Ai-Ling Lin5, and Priti Balchandani1 | ||
1Biomedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2Computer Science, University of Kentucky, Lexington, KY, United States, 3Psychiatry, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 4Diagnostic, Molecular and Interventional Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 5Radiology, University of Missouri, Columbia, MO, United States |
||
Major depression is highly-prevalent disorder with frustratingly-high rates of treatment resistance. Ultrahigh field imaging may provide objective quantitative biomarkers for characterizing depression, generating insight into clinical phenotypes of this heterogeneous disease. Forty-two major depressive disorder patients currently off anti-depressant treatment were recruited for scanning at ultrahigh field, and given batteries of clinical symptom measures. Machine-learning clustering analysis was performed to group patients by clinical symptoms and differences in imaging features observed. A separate analysis was performed in the reverse direction clustering on quantified imaging features and identifying clinical differences between clusters, including differences in ruminative response between the patient clusters. |
| 1518 | Computer 68
|
Towards Clinical Translation of Machine Learning-based Automated Prescription of Spine MRI Acquisitions |
| Eugene Ozhinsky1, Felix Liu1, Valentina Pedoia1, and Sharmila Majumdar1 | ||
1Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States |
||
High quality scan prescription that optimally covers the area of interest with scan planes aligned to relevant anatomical structures is crucial for error-free radiologic interpretation. In this study we used images and metadata from previously acquired examinations of lumbar spine to train machine learning-based automated prescription models without the need of any manual annotation or feature engineering. The automated prescription pipeline was integrated with the scanner console software and evaluated in healthy volunteer experiments. This study demonstrates the feasibility of using oriented object detection-based pipelines on the scanner for automated prescription of lumbar spine acquisitions. |
| 1519 | Computer 69
|
Comparison of whole-prostate radiomics models of disease severity derived from expert and AI based prostate segmentations |
| Paul E Summers1, Lars Johannes Isaksson2, Matteo Johannes Pepa2, Mattia Zaffaroni2, Maria Giulia Vincini2, Giulia Corrao2,3, Giovanni Carlo Mazzola2,3, Marco Rotondi2,3, Sara Raimondi4, Sara Gandini4, Stefania Volpe2,3, Zaharudin Haron5, Sarah Alessi1, Paola Pricolo1, Francesco Alessandro Mistretta6, Stefano Luzzago6, Federico Cattani7, Gennaro Musi3,6, Ottavio De Cobelli3,6, Marta Cremonesi8, Roberto Orecchia9, Giulia Marvaso2,3, Barbara Alicja Jereczek-Fossa2,3, and Giuseppe Petralia3,10 | ||
1Division of Radiology, IEO, European Institute of Oncology IRCCS, Milano, Italy, 2Division of Radiation Oncology, IEO, European Institute of Oncology IRCCS, Milano, Italy, 3Department of Oncology and Hemato-oncology, University of Milan, Milano, Italy, 4Department of Experimental Oncology, IEO, European Institute of Oncology IRCCS, Milano, Italy, 5Radiology Department, National Cancer Institute, Putrajaya, Malaysia, 6Division of Urology, IEO, European Institute of Oncology IRCCS, Milano, Italy, 7Unit of Medical Physics, IEO, European Institute of Oncology IRCCS, Milano, Italy, 8Radiation Research Unit, IEO, European Institute of Oncology IRCCS, Milano, Italy, 9Scientific Directorate, IEO, European Institute of Oncology IRCCS, Milano, Italy, 10Precision Imaging and Research Unit, IEO, European Institute of Oncology IRCCS, Milano, Italy |
||
A persisting concern is that downstream models of clinical endpoints may depend on whether the contours were drawn by an expert or an AI. Prediction models for surgical margin status, and pathology-based lymph nodes, tumor stage and ISUP grade group were formed using clinical and radiological features along with whole-prostate radiomic features based on manual and AI segmentations of the prostate in 100 patients who proceeded to prostatectomy after multiparametric-MRI. The models based on AI segmented prostates differed from those based on manual segmentation, but with similar if not better performance. Further testing of generalizability of the models is required. |
| 1520 | Computer 70
|
A Deep Neural Network for Detection of Glioblastomas in Spectroscopic MRI |
| Erin Beate Bjørkeli1,2, Jonn Terje Geitung1,2, and Morteza Esmaeili1,3 | ||
1Department of Diagnostic Imaging, Akershus University Hospital, Oslo, Norway, 2Institue of Clinical Medicine, University of Oslo, Oslo, Norway, 3Department of Electrical Engineering and Computer Science, University of Stavanger, Stavanger, Norway |
||
We have developed a MRSI spectra classifying convolutional neural network (CNN), building on the AUTOMAP model for image and spectra reconstruction. The model was trained to discern between non-water-suppressed spectra from healthy subjects and glioblastoma (GBM) patients. The trained CNN was able to classify the spectra correctly and seemed to recognize the healthy spectra based on the NAA-peak and the GBM based on the choline levels and possibly 2HG, indicating an IDH mutation. |
| 1521 | Computer 71
|
Predicted tumor stroma segmentation from high-field MR texture maps and machine learning: an ex vivo study on ovarian tumors |
| Marion Tardieu1, Lakhdar Khellaf2, Maida Cardoso3, Olivia Sgarbura4, Pierre-Emmanuel Colombo4, Christophe Goze-Bac3, and Stephanie Nougaret1 | ||
1Montpellier Cancer Research Institute (IRCM), INSERM U1194, University of Montpellier, Montpellier, France, 2Department of pathology, Montpellier Cancer Institute (ICM), Montpellier, France, 3BNIF facility, L2C, UMR 5221, CNRS, University of Montpellier, Montpellier, France, 4Department of Surgery, Montpellier Cancer Institute (ICM), Montpellier, France |
||
The objective was to probe the associations of high-field MR-images and their derived texture maps (TM) with histopathology in ovarian cancer (OC). Four ovarian tumors were imaged ex-vivo using a 9.4T-MR scanner. Automated MR-derived stroma-tumor segmentation maps were constructed using machine learning and validated against histology. Through TM, we found that areas of tumor cells appeared uniform on MR-images, while areas of stroma appeared heterogeneous. Using the automated model, MRI predicted stromal proportion with an accuracy from 61.4% to 71.9%. In this hypothesis-generating study, we showed that it is feasible to resolve histologic structures in OC using ex-vivo MR radiomics. |
| 1522 | Computer 72
|
Virtual Dynamic Contrast Enhanced MRI of the Breast using a U-Net |
| Hannes Schreiter1,2, Vishal Sukumar1,2, Lorenz Kapsner1, Lukas Folle2, Sabine Ohlmeyer1, Frederik Bernd Laun1, Evelyn Wenkel1, Michael Uder1, Andreas Maier2, Sebastian Bickelhaupt1, and Andrzej Liebert1 | ||
1Institute of Radiology, University Hospital Erlangen, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Erlangen, Germany, 2Patter Recognition Lab, Department of Computer Science, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Erlangen, Germany |
||
Magnetic resonance imaging (MRI) examinations of the breast require intravenous administration of gadolinium based contrast agents (GBCA) for comprehensive characterization of the tissue. Novel approaches reducing the need for GBCA might therefore be of value. Here a virtual dynamic contrast enhancement (vDCE) using a U-net architecture is investigated in a cohort of n=540 patients. The vDCE generates T1 subtraction images for five consecutive time points predicting the perfusion maps based on native T1-weighted, T2-weighted, and multi-b-value diffusion weighted acquisitions. A mean structural similarity index (SSIM) value over a test group of 82 patients of 0.848±0.025 was achieved. |
| 1603 | Computer 58
|
GRASP-Pro+: GRASP reconstruction with locally low-rank subspace constraint for DCE-MRI |
| Eddy Solomon1, Jonghyun Bae1,2, Elcin Zan2, Linda Moy2, Yulin Ge2, Li Feng3, and Gene Sungheon Kim1 | ||
1Department of Radiology, MRI Research Institute, Weill Cornell Medicine, New York, NY, United States, 2Department of Radiology, New York University, New York, NY, United States, 3BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States |
||
While the globally low-rank (GLR) model has been demonstrated to be effective in representing global contrast change, it is expected to be less effective for spatially localized signal dynamics. In this work, we propose an improved reconstruction framework, which extends GRASP-Pro using a locally low-rank (LLR) model to represent spatially localized dynamics based on clusters or patches information. This approach has been tested in multiple DCE applications including both cancer and healthy subjects. In addition, we propose an anatomical cluster-based reconstruction approach for brain DCE-MRI. |
| 1604 | Computer 59
|
Rapid CS-Wave MPRAGE acquisition with automated parameter selection |
| Gabriel Varela-Mattatall1,2,3, Tae Hyung Kim3, Jaejin Cho3, Wei-Ching Lo4, Borjan A. Gagoski5,6, Ravi S. Menon1,2, and Berkin Bilgic3,7 | ||
1Centre for Functional and Metabolic Mapping (CFMM) | Robarts Research Institute | Western University, London, ON, Canada, 2Department of Medical Biophysics | Schulich School of Medicine and Dentistry | Western University, London, ON, Canada, 3Department of Radiology | Athinoula A. Martinos Center for Biomedical imaging | Massachusetts General Hospital & Harvard Medical School, Charlestown, MA, United States, 4Siemens Medical Solutions USA, Inc., Charlestown, MA, United States, 5Department of Radiology | Harvard Medical School, Boston, MA, United States, 6Fetal-Neonatal Neuroimaging & Developmental Science Center | Boston Children's Hospital, Boston, MA, United States, 7Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States |
||
Wave encoding mitigates g-factor noise amplification in highly accelerated parallel imaging but achieving ultra-high acceleration factors is precluded by the intrinsic “√R” SNR penalty. To overcome this limitation, we propose a compressed sensing-based reconstruction with automatic selection of the regularization weighting. Moreover, we show that CS-Wave is flexible enough to perform well with uniform undersampling. We compare reconstruction performance of CS-Wave against the state-of-art Wave-LORAKS which requires parameter tuning, and evaluate different undersampling patterns at R=12-fold acceleration. Results indicate higher reconstruction quality and showcase the feasibility of ultra-fast Wave-MPRAGE acquisitions. |
| 1605 | Computer 60
|
EigenGRASP: Subject-Specific Temporal-Learning Radial Sampling Image Reconstruction in Dynamic Contrast-Enhanced MRI |
| Ramin Jafari1, Masoud Zarepisheh1, Richard Kinh Gian Do2, and Ricardo Otazo1 | ||
1Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2Memorial Sloan Kettering Cancer Center, New york, NY, United States |
||
GRASP is an image reconstruction algorithm for free-breathing dynamic contrast-enhanced MRI which uses universal L1-type regularization to suppress undersampling artifacts. We propose to replace it with a subject-specific data-driven L2-type regularization which can improve image quality and decrease reconstruction time. |
| 1606 | Computer 61
|
A Locally Structured Low-Rank Tensor Method Using Submatrix Constraints for Joint Multi-echo Image Reconstruction |
| Xi Chen1, Wenchuan Wu1, and Mark Chiew1 | ||
1Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom |
||
Recently, we proposed an improved structured low-rank (SLR) reconstruction - locally structured low-rank (LSLR) method, which forces low-rank constraints on submatrices of the Hankel structured matrix. This simple modification leads to a robust improvement over conventional SLR reconstruction, which has been validated on the low-rank tensor reconstruction for the multi-echo GRE data. |
| 1607 | Computer 62
|
Applying L+S decomposition to improve sensitivity to dynamic changes in functional MR spectroscopy |
| Adam Berrington1 and I. Betina Ip2 | ||
1Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, United Kingdom, 2Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom |
||
Functional MRS (fMRS) is a powerful technique to measure metabolite responses over time. However, noise and spectral contamination limit the ability to study individual metabolite time-courses. In this work, we propose to model fMRS spectra as a superposition of low-rank (L) and sparse components (S). L+S decomposition resulted in separation of temporally-correlated signal from noise in simulation. In vivo, L+S spectra had higher SNR compared to original data (P=0.007) and the mean glutamate time-course, using L+S spectra, was more strongly correlated to stimulus. L+S decomposition is a promising data-driven method to enhance sensitivity to dynamic changes in fMRS. |
| 1608 | Computer 63
|
Joint reconstruction of multi-TE diffusion MRI acquired using TDM-EPI with complementary k-space sampling |
| Yang Ji1,2, Congyu Liao3, William Scott Hoge4, Berkin Bilgic5, Yogesh Rathi1,4, and Lipeng Ning1 | ||
1Department of Psychiatry, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States, 2Wellcome Centre for Integrative Neuroimaging, FMRIB Division, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States, 5Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States |
||
Combined diffusion-relaxometry has demonstrated promising capability to noninvasively probe tissue microstructure by joint modeling of relaxation coefficients and diffusivity. Our recent work has introduced a sequence based on the time-division multiplexing technique to accelerate the acquisition of relaxation-diffusion MRI. In this work, we further developed the TDM-EPI sequence by integrating ky-shifted k-space sampling strategies for data acquired at different TEs. Moreover, we implemented and compared several reconstruction methods to integrate complementary k-space samples to joint estimate images at different TEs. The results showed that the joint reconstruction approach can improve image quality and reduce artifact compared with conventional reconstruction methods. |
| 1609 | Computer 64
|
Pitfalls of data-driven tSNR optimized coil combination for fMRI |
| Redouane Jamil1, Franck Mauconduit1, Caroline Le Ster1, Philipp Ehses2, Benedikt A Poser3, Alexandre Vignaud1, and Nicolas Boulant1 | ||
1CEA, CNRS, BAOBAB, NeuroSpin, Gif-sur-Yvette, France, France, 2German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, 3Department of Cognitive Neuroscience, Maastricht Brain Imaging Centre, Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, Netherlands |
||
For MRI with a multi-receiver RF coil, one image per coil element and per time frame is obtained. The final image is typically calculated from the root sum of squares (rSoS) combination across channels. While this combination approach is quasi-optimal for SNR, it is not necessarily optimal for temporal SNR (tSNR) of the time-series. We present two analytical and voxel-wise coil combination expressions reaching optimality in tSNR and t-score for the mean (TSM) respectively. Their BOLD sensitivity is compared to the gold standard covariance root sum of squares. Both improved tSNR and TSM but yielded weaker t-scores than covSoS. |
| 1610 | Computer 65
|
Fast 3D DCE MRI using partially separable model: autopilot Multitasking |
| Jingyuan Lyu1, Qi Liu1, Yu Ding1, Jian Xu1, Xiaomao Gong2, Weiguo Zhang1, Shengxiang Rao3,4,5, Wentao Wang3,4,5, and Mengsu Zeng3,4,5 | ||
1UIH America, Inc., Houston, TX, United States, 2United Imaging Healthcare, Shanghai, China, 3Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China, 4Department of Medical Imaging, Shanghai Medical College, Fudan University, Shanghai, China, 5Shanghai Institute of Medical Imaging, Shanghai, China |
||
Inspired by Multitasking, we have developed a fast reconstruction method for 3D stack-of-star (SoS) acquisitions to achieve high temporal resolution (0.1s) DCE MRI without additional navigator data. The proposed method is fully compatible with existing sequence using 3D SoS trajectories, but with less computation complexity. |
| 1611 | Computer 66
|
MORE-SPARKLING: Non-Cartesian trajectories with Minimized Off-Resonance Effects |
| Chaithya G R1,2, Guillaume Daval-Frérot1,2,3, Aurélien Massire3, Boris Mailhe4, Mariappan Nadar4, Alexandre Vignaud1, and Philippe Ciuciu1,2 | ||
1NeuroSpin, Joliot, CEA, Université Paris-Saclay, F-91191, Gif-sur-Yvette, France, 2Inria, Parietal, Université Paris-Saclay, F-91120, Palaiseau, France, 3Siemens Healthcare SAS, Saint-Denis, 93210, France, 4Siemens Healthineers, Princeton, NJ, United States |
||
We augment the recently introduced SPARKLING algorithm and propose an improved mathematical formulation that takes the temporal dependence of the MR signal into account. This prevents the trajectories from sampling similar portions of k-space at different times, thereby reducing distortions and blurring induced by \(B_0\) inhomogenieties. Overall, these trajectories present a smooth distribution over time in k-space and Minimized Off-Resonance Effects (MORE-SPARKLING), verified both retrospectively and prospectively with scans performed in vivo at 3T on a healthy volunteer. |
| 1612 | Computer 67
|
Spiral 3DREAM sequence for fast whole-brain B1 Mapping |
| Marten Veldmann1, Svenja Niesen1, Philipp Ehses1, and Tony Stöcker1,2 | ||
1Deutsches Zentrum für Neurodegenerative Erkrankungen e.V. (DZNE), Bonn, Germany, 2Department of Physics & Astronomy, University of Bonn, Bonn, Germany |
||
The 3DREAM sequence provides a fast technique for whole-brain B1 mapping, but suffers from different blurring levels in FID and STE* images. We developed a new variant of the 3DREAM sequence with a spiral readout to mitigate this problem. The spiral readout shortens the echo train duration and reduces the number of excitations after the STEAM preparation significantly. This leads to similar blurring levels in FID and STE* images and increases the effective resolution of flip angle maps. As a result, correction strategies for different blurring levels are no longer necessary. |
| 1613 | Computer 68
|
Evaluation of 3D SPARKLING readout for Sodium UTE MRI at ultra-high magnetic field |
| Renata Porciuncula Baptista1, Alexandre Vignaud1, Chaithya G R1,2, Guillaume Daval-Frérot1,2, Franck Mauconduit1, Mathieu Naudin3, Marc Lapert4, Remy Guillevin3, Philippe Ciuciu1,2, Cécile Rabrait-Lerman1, and Fawzi Boumezbeur1 | ||
1NeuroSpin, Joliot, CEA, CNRS, University Paris-Saclay, Gif-sur-Yvette, France, 2Inria, Parietal, Université Paris-Saclay, Palaiseau, France, 3University Hospital Centre Poitiers, DACTIM-MIS, Poitiers, France, 4Siemens Healthcare SAS, Saint-Denis, France |
||
Quantitative 23Na MRI provides useful information about brain tissue homeostasis. Most sequences use deterministic non-Cartesian 3D trajectories such as TPI. However, stochastic strategies such as SPARKLING could improve the coverage of k-space. This study evaluates the advantages of SPARKLING versus TPI. From in vivo datasets at 7T, we determined that undersampled SPARKLING acquisitions outperform TPI for (8 mm)3 resolution with a birdcage coil or for (4 mm)3 with a 32-channel coil. Through extrapolation of these results, we predict that at 11.7T/32-channel, 23Na MRI data could be acquired in 90s at (3 mm)3, which could be interesting for sodium fMRI imaging. |
| 1614 | Computer 69
|
SNR Enhancement of MR Images with Variable Compress Sensed Excitation: vNEX |
| Harsh Kumar Agarwal1, Shaik Ahmed1, Rajagopalan Sundaresan1, Sudhanya Chatterjee1, Sajith Rajamani1, Ashok Kumar P Reddy1, Bhairav Mehta1, Dattesh Shanbhag1, and Ramesh Venkatesan1 | ||
1GE Healthcare, Bangalore, India |
||
MR imaging is signal starve leading to long acquisition time. Multiple averaging/excitation is most common way to boost the SNR. However, it is associated with proportionately long scan time. This abstract presents Variable Compress Sensed Excitation (vNEX) image acquisition and image reconstruction technique which subsample each signal average by compressed sensing fast MRI technique and robust image reconstruction is done to reconstruct MRI images for individual signal averages. The proposed technique is demonstrated for T2w brain MRI where SNR equivalence of 336sec scan time is shown in a 200sec vNEX MRI. Three sampling patterns were proposed and statistically compared. |
| 1615 | Computer 70
|
MRI with Sub-Millisecond Temporal Resolution: SPIRE Imaging of a High-speed Motion Phantom |
| Fischer Johannes1 and Michael Bock1 | ||
1Dept. of Radiology, Medical Physics, Medical Center University of Freiburg, Faculty of Medicine, Uni, Freiburg, Germany |
||
Imaging sequences with sub-millisecond temporal resolution based on single point imaging (SPI) are very time inefficient. We designed and imaged a motion phantom capable of velocities up to 5.5m/s. A synchronization signal is acquired using a photoelectric barrier and combined with a sequence trigger signal. Reconstructed images show clear delineation of small structures with a spatial resolution of $$$\Delta x$$$=0.67mm and a temporal resolution of $$$\delta t$$$=246μs. |
| 1616 | Computer 71
|
Spiral UTE MRI of the Lung: An Investigation on Motion Sensitivity |
| Valentin Fauveau1, Adam Jacobi2, Adam Bernheim2, Michael Chung2, Yang Yang1,2, Thomas Benkert3, Zahi Fayad1,2, and Li Feng1,2 | ||
1BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 3MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany |
||
Spiral UTE is a relatively new MRI technique that combines ultra-short echo time acquisition with a stack-of-spirals trajectory for imaging the lung. In this work, we aimed to analyze the motion sensitivity of different reordering schemes and breath holding positions in spiral UTE MRI to determine the optimized protocol for imaging the lungs. A blind assessment was performed by three experienced chest radiologists and the results were analyzed with statistical tests. The Spiral UTE with line-in-partition reordering performed during the inspiration phase was considered the best protocol in our study. |
| 1617 | Computer 72
|
SPEN, RESOLVE and Spin-Echo EPI at 7T: A single-Tx human brain scan comparison |
| Martins Otikovs1 and Lucio Frydman1,2 | ||
1Department of Chemical and Biological Physics, Weizmann Institute of Science, Rehovot, Israel, 2Azrieli National Center for Brain Imaging, Weizmann Institute of Science, Rehovot, Israel |
||
Spatiotemporal encoding (SPEN) is an ultrafast imaging technique, that can serve as useful alternative to single shot and to read-out segmented (RESOLVE) spin-echo (SE) EPI methods, when trying to overcome image distortions due to magnetic field (B0) inhomogeneities. SPEN’s reliance on adiabatic sweeps could also make it an attractive candidate for imaging at high fields, and overcoming B1+-related inhomogeneities. Herein we present phantom and in-vivo results acquired on a 7T human imaging system for single and multi-shot SPEN acquisitions, corroborating its robustness vs comparable single and multi-shot SE-EPI and RESOLVE acquisitions implemented on a single-Tx configuration. |
| 1618 | Computer 73
|
A vendor-neutral 3D MP-RAGE implementation in Pulseq for reduced inter-site variability |
| Yogesh Rathi1, Jon-Fredrik Nielsen2, Maxim Zaitsev3, Lipeng Ning1, Jr-Yuan Chiou4, Tae Hyung Kim5, William Grissom6, Borjan Gagoski7, and Berkin Bilgic5 | ||
1Psychiatry, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States, 2University of Michigan, Ann Arbor, MI, United States, 3High Field MR Center, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria, 4Brigham and Women's Hospital, Boston, MA, United States, 5Athinoula A. Martinos Center for Biomedical Imaging, Harvard Medical School, Boston, MA, United States, 6Vanderbilt University, Nashville, TN, United States, 7Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital, Harvard Medical School, Boston, MA, United States |
||
Inter-scanner variability can reduce the statistical power of neuroimaging studies due to the increased variance in the data. While existing efforts employ consistent high level parameters to acquire data across scanners, we hypothesize that differences due to sequence implementation and reconstruction algorithms contribute substantially to variability in multi-center settings. We propose a unified vendor-neutral MP-RAGE protocol based on the Pulseq sequence development platform that can be used consistently across sites and vendors for reducing inter-scanner variability. Preliminary results show increased consistency of cortical thickness across sites using Pulseq protocol compared to vendor-provided sequences. |
| 1619 | Computer 74
|
Fast and motion robust 2D T2 TSE Propeller acquisition of the prostate with Compressed SENSE: comparison with conventional SENSE acquisition |
| Leon Bischoff1, Christoph Katemann2, Oliver Weber2, Alexander Isaak1, Dmitrij Kravchenko1, Narine Mesropyan1, Christoph Endler1, Thomas Vollbrecht1, Claus Christian Pieper1, Ulrike Attenberger1, and Julian Luetkens1 | ||
1Department of Diagnostic and Interventional Radiology, University Hospital Bonn, Bonn, Germany, 2Philips GmbH Market DACH, Hamburg, Germany |
||
Multiparametric MRI (mpMRI) of the prostate can detect clinically significant prostate cancer in men. To accelerate the time-consuming acquisition process, we integrated a new Compressed SENSE (CS) method for T2-weighted sequences with propeller acquisition and compared it qualitatively and quantitatively to conventional SENSE accelerated T2-weighted propeller sequences. We could demonstrate that while the new CS acquisition method has less artifacts, better image sharpness and higher apparent signal-to-noise ratio (aSNR) and contrast-to-noise ratio (CNR), it reduced the acquisition time by 24%. These findings indicate a superiority over the conventional T2-sequence and could improve the diagnostic workup of patients with prostate cancer. |
| 1620 | Computer 75
|
Imaging Off-Center: How far can we move outside the central imaging region |
| Sebastian Littin1, Patrick Hucker1, Maximilian Frederik Russe2, Feng Jia1, Philipp Amrein1, and Maxim Zaitsev1,3 | ||
1Department of Radiology, Medical Physics, University Freiburg, Faculty of Medicin, Freiburg, Germany, 2Department of Radiology, University Freiburg, Faculty of Medicin, Freiburg, Germany, 3Center for Medical Physics and Biomedical Engineering, High Field MR Center, Medical University of Vienna, Vienna, Austria |
||
Imaging outside the predefined and optimized FoV may enhance accessibility for interventional MRI procedures. Here we present different measures to assess the encoding capability of gradient systems in off-center regions. Different clinically establishes sequences are compared at 0.55T and 1.5T regarding their imaging capabilities beyond the predefined target regions. |
| 1621 | Computer 76
|
Feasibility of 3D Adiabatic T1ρ-prepared Fast Spin Echo (3D Adiab-T1ρ-FSE) Imaging |
| Hyungseok Jang1, Yajun Ma1, Dina Moazamian1, Michael Carl2, Saeed Jerban1, Alecio F Lombardi1, Christine B Chung1,3, Eric Y Chang1,3, and Jiang Du1 | ||
1Radiology, University of California San Diego, San Diego, CA, United States, 2GE Healthcare, San Diego, CA, United States, 3Radiology Service, Veterans Affairs San Diego Healthcare System, San Diego, CA, United States |
||
T1ρ has been investigated as a quantitative biomarker sensitive to changes in macromolecules such as proteoglycan and collagen in musculoskeletal systems. More recently, adiabatic T1ρ (Adiab-T1ρ) has emerged as an alternative to conventional continuous wave T1ρ to reduce the magic angle effect, which is a major confounding factor in accurate T1ρ estimation. In this study, we investigated a new pulse sequence combining adiabatic T1ρ preparation and efficient 3D fast spin echo (FSE) for more robust adiab-T1ρ mapping in the human knee. The efficacies of RF cycling and magnetization reset were also demonstrated. |
| 1696 | Computer 65
|
Correcting for gradient non-linearity in concurrent field monitoring |
| Paul I Dubovan1,2, Kyle M Gilbert1,2, and Corey A Baron1,2 | ||
1Medical Biophysics, Western University, London, ON, Canada, 2Centre for Functional and Metabolic Mapping, Western University, London, ON, Canada |
||
Concurrent field monitoring (FM) via integration of field probes in a radiofrequency coil provides advantages over sequential field monitoring that include patient-specific corrections, as well as improvements to user workflow. However, specific design considerations can require placing field probes far from isocentre where gradient fields are no longer linear, reducing data integrity when fitting high order spatially varying terms. We propose a three-part correction algorithm that seeks to correct these errors and compare FM data, as well as reconstructed images, before and after correction. Correction improved integrity of FM data and enhanced quality of anatomical and diffusion weighted images. |
| 1697 | Computer 66
|
A Beamforming-Based Coil Combination Method to Reduce Streaking Artifacts and Preserve Phase Fidelity in Radial MRI |
| Shu-Fu Shih1,2 and Holden H. Wu1,2 | ||
1Department of Radiological Sciences, University of California Los Angeles, Los Angeles, CA, United States, 2Department of Bioengineering, University of California Los Angeles, Los Angeles, CA, United States |
||
Radial acquisition can be sensitive to hardware imperfections such as gradient non-linearity and B0 field inhomogeneity. This usually becomes accentuated in areas more distant to the isocenter and can lead to streaking artifacts even when nominally fulfilling Nyquist criteria. Previous work demonstrated beamforming-based methods for radial streaking reduction, but did not explicitly consider phase and did not evaluate the artifact-suppression performance on phase explicitly. In this work, we developed a new beamforming formulation based on minimum-variance distortionless response (MVDR) that can suppress streaking artifact while preserving accurate phase information |
| 1698 | Computer 67
|
Fast Multi-Parametric Mapping with an Extended Spiral and Off-Resonance Readout Correction |
| Willem van Valenberg1, Gyula Kotek1, Mika W. Vogel2, Juan A. Hernandez-Tamames1, and Dirk H. J. Poot1 | ||
1Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands, 2GE Healthcare, Hoevelaken, Netherlands |
||
Single-shot quantitative methods with spiral sampling can increase the readout duration to reduce undersampling artifacts in the images and their propagation to the parameter maps. However, a longer readout increases blurring due to the off-resonance field if not accounted for in the reconstruction. We propose an efficient reconstruction that 1. corrects for blurring due to the off-resonance field, and 2. reduces undersampling artifacts by constraining the signal variation in each voxel to a predetermined subspace. The reconstruction method is shown to enable accurate multi-parametric mapping of a single-shot MP-b-nSSFP acquisition with submillimeter resolution in 1.8s. |
| 1699 | Computer 68
|
BPE XD-GRASP: Using GROG-BPE for improved respiratory motion compensation in dynamic contrast enhanced golden-angle radial MRI |
| Yumna Bilal1,2, Ibtisam Aslam1,3, Muhammad Faisal Siddiqui1, and Hammad Omer1 | ||
1Medical Image Processing Research Group (MIPRG), Department of Electrical & Computer Engineering, COMSATS University Islamabad, Islamabad, Pakistan, 2Department of Electrical Engineering, University of Gujrat, GUJRAT, Pakistan, 3Service of Radiology, Geneva University Hospitals and Faculty of Medicine, University of Geneva, Geneva, Switzerland |
||
This work aims at incorporating the GROG facilitated Bunch Phase Encoding (BPE) in XD-GRASP reconstruction framework for improved motion compensation in free-breathing Dynamic Contrast Enhanced golden-angle radial MRI data. In the proposed method, the inherent randomness and added redundancy in the BPE data, is exploited by the Compressed Sensing (CS) algorithm resulting in images with higher quality and improved mitigation of respiratory motion artifacts as compared to the conventionally used XD-GRASP method. |
| 1700 | Computer 69
|
Sensitivity Analysis of the Bloch Equations |
| Nick Scholand1,2,3 and Martin Uecker1,2,3 | ||
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria, 2Department of Interventional and Diagnostic Radiology, University Medical Center Göttingen, Göttingen, Germany, 3German Centre for Cardiovascular Research, Göttingen, Germany |
||
Algorithms for pulse optimization, quantitative MRI, or model-based reconstruction require knowledge about partial derivatives of the Bloch equations.While being available for specific analytical solutions, computing them becomes challenging in the general case. Difference quotient techniques can be used, but require perturbation tuning to avoid errors. In this work, we investigate the use of direct sensitivity analysis to estimate the partial derivatives of the Bloch equations. We validate it with an analytical sequence model and compare it to the difference quotient in an example without analytical solution. In all cases, direct sensitivity analysis provided highly accurate estimates of the partial derivatives. |
| 1701 | Computer 70
|
Reproducibility of IVIM Quantification Across Diffusion Gradient Waveforms using Pseudo-Diffusion and Physical IVIM Signal Models |
| Gregory Simchick1,2 and Diego Hernando1,2 | ||
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, 2Medical Physics, University of Wisconsin-Madison, Madison, WI, United States |
||
The intravoxel incoherent motion (IVIM) DWI signal has been shown to depend on both b-value and the first motion moment (M1). However, traditional pseudo-diffusion IVIM signal models only include b-value dependence. In this work, monopolar and 2D (b-M1) IVIM acquisitions were acquired using diffusion gradient waveforms with different maximum amplitudes to emulate different gradient hardware. From each acquisition, IVIM estimates were obtained by fitting pseudo-diffusion and physical (M1 dependent) IVIM signal models. Estimated IVIM parameters were compared across acquisitions. Reproducible physical IVIM estimates were obtained using 2D IVIM-DWI acquisitions, while pseudo-diffusion IVIM estimates obtained using monopolar acquisitions demonstrated poor reproducibility. |
| 1702 | Computer 71
|
Quantification of grey matter sodium content in joint sodium MR reconstructions using anatomical regularization and T2* estimation at 3T |
| Georg Schramm1, Yongxian Qian2, Johan Nuyts1, and Fernando Boada2 | ||
1KU Leuven, Leuven, Belgium, 2New York University School of Medicine, New York City, NY, United States |
||
Sodium MR signal acquisition and image reconstruction are severely hampered by the fast relaxation properties of the sodium nuclei. Signal decay due to fast transverse relaxation during readout leads to loss of high frequency information and severe partial volume effects (e.g.signal spill over from CSF into GM). In this work, we apply a reconstruction framework including anatomical regularization and signal decay modeling and estimation to dual echo sodium data to three subjects to investigate its effect on cortical grey matter sodium quantification. |
| 1703 | Computer 72
|
Evaluating the impact of respiratory binning strategies on 4D-MRI reconstruction for an MR-Linac |
| Bastien Lecoeur1, Marco Barbone2, Sophie Alexander3, Jessica Gough3, Uwe Oelfke1, Wayne Luk2, and Andreas Wetscherek1 | ||
1Joint Department of Physics, The Institute of Cancer Research, London, United Kingdom, 2Department of Computing, Imperial College London, London, United Kingdom, 3The Royal Marsden NHS Fundation Trust, London, United Kingdom |
||
In the context of MR-guided radiotherapy, 4D-MRI is of particular interest for lung and abdominal cancer treatment, as it enables quantifying the extent of respiratory motion at the time of treatment, facilitating time-efficient midposition treatments. To reduce long reconstruction times of iterative compressed sensing-based reconstructions involving algorithms, such as XD-GRASP, we used a fast C++ implementation. We evaluated the impact of using overlapping respiratory bins and different self-gating signals on image quality and reconstruction time in multiple patients with and without abdominal compression belts. |
| 1704 | Computer 73
|
A Fourier representation of the diffusion MRI signal |
| Chengran Fang1, Demian Wassermann1, and Jing-Rebecca Li1 | ||
1INRIA Saclay, PALAISEAU, France |
||
The diffusion MRI community lacks a spectral perspective to investigate the intricate relationship between the diffusion MRI signals and cellular structures. Our study proposes a new simulation method that can provide a Fourier representation of the signal. Common spectral methods suffer from the singularity of the heat kernel, which requires infinite Fourier modes. We overcome the singularity by formulating the solution in a convolutional form and splitting the heat kernel into a singular part and a smooth part. Numerical experiments are performed to demonstrate the convergence of our method. |
| 1705 | Computer 74
|
Convex Optimized Diffusion Encoding (CODE) with Partial Fourier Imaging for EPI Diffusion Weighted Imaging |
| Matthew J. Middione1, Tyler E. Cork1,2, Michael Loecher1, Tim Sprenger3, Arnaud Guidon4, Shreyas S. Vasanawala1, and Daniel B. Ennis1 | ||
1Department of Radiology, Stanford University, Stanford, CA, United States, 2Department of Bioengineering, Stanford University, Stanford, CA, United States, 3GE Healthcare, Munich, Germany, 4GE Healthcare, Boston, MA, United States |
||
Diffusion weighted imaging typically uses monopolar (MONO) diffusion gradient waveforms, which may have sequence dead time that extends the TE and reduces the SNR. Partial Fourier (PF) imaging is routinely used in DWI to shorten the TE (improving SNR), but can lead to image blurring. Convex Optimized Diffusion Encoding (CODE) is a constrained optimization technique that designs time-optimal diffusion encoding gradient waveforms without sequence dead time. CODE produces a shorter TE than MONO, which leads to increased SNR. Herein, we show that CODE without PF has increased SNR and reduced image blurring compared to MONO with PF=6/8. |
| 1706 | Computer 75
|
Protocol Optimization of Spectro-Dynamic MRI |
| Max H.C. van Riel1, Niek R.F. Huttinga1, Tom Bruijnen1, and Alessandro Sbrizzi1 | ||
1Department of Radiotherapy, Computational Imaging Group for MR diagnostics and therapy, UMC Utrecht, Utrecht, Netherlands |
||
Inferring dynamical information from (bio)mechanical systems at a high temporal resolution can be very valuable for cardiovascular or musculoskeletal applications. Spectro-dynamic MRI is a recently proposed method that combines a measurement model and a dynamical model to characterize dynamical systems directly from k-space data. Both the displacement fields and the underlying dynamical parameters are estimated. In this work, different sampling trajectories and acquisition orderings are used to investigate the trade-off between temporal resolution and k-space coverage. A phantom experiment shows that it is possible to reconstruct a moving image from the estimated dynamics at a temporal resolution of 4.4 ms. |
| 1707 | Computer 76
|
Sampling Mask Optimization for Compressed Sensing PETRA MRI |
| Serhat Ilbey1, Ali Özen1, Pia Jungmann2,3, Johannes Fischer1, Matthias Jung2, and Michael Bock1 | ||
1Dept.of Radiology, Medical Physics, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 2Department of Diagnostic and Interventional Radiology, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 3Department of Radiology, Cantonal Hospital Grisons, Chur, Switzerland |
||
Single point imaging (SPI) is intrinsically time-inefficient resulting in very long scan times in PETRA. Compressed sensing PETRA (csPETRA) shortens the acquisition time – here, various csPETRA undersampling schemes were generated and an optimal undersampling scheme was determined experimentally. It was shown in csPETRA knee imaging that in Poisson sampling combined with a spherical fully-sampled center results in minimum error compared to the fully-sampled acquisitions. |
| 1708 | Computer 77
|
EPIFANI: an ultrafast T1, B1 and magnetization mapping technique |
| Marco Andrea Zampini1,2, Jan Sijbers3, Marleen Verhoye2, and Ruslan Garipov1 | ||
1MR Solutions Ltd, Guildford, United Kingdom, 2Department of Biomedical Sciences, University of Antwerp, Wilrijk, Belgium, 3Department of Physics, University of Antwerp, Wilrijk, Belgium |
||
Quantitative MRI still struggles to be used in clinical practice due to long acquisition times. We introduce EPIFANI, an ultrafast B1-corrected T1 mapping technique which uses EPI images acquired in two successive repetition times and a parametric map fit used in VAFI. Both a single-shot and a multi-shot version of EPIFANI were tested on phantoms and in vivo and report accurate T1 values. |
| 1709 | Computer 78
|
Rapid prototyping of a motion-robust 2D Radial GRE sequence with reduced acoustic noise generation using Pulseq |
| Elisa Marchetto1,2, Maxim Zaitsev3, and Daniel Gallichan1,2 | ||
1School of Engineering, Cardiff University, Cardiff, United Kingdom, 2Cardiff University Brain Research Centre (CUBRIC), Cardiff University, Cardiff, United Kingdom, 3High Field MR Center, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria |
||
A radial GRE-sequence was modified using Pulseq to reduce the acoustic noise amount produced by the rapid switching of gradient currents without excessively increasing the TR-time. Prospective motion-correction was performed using a markerless tracking device that sent the motion estimates to the scanner via the libXPACE framework, incorporated directly into Pulseq, to update gradients and RF pulses during the MR scan. The quieter sequence resulted in a acoustic noise reduction of 6 dB(A) (i.e. a factor of 2) compared to the standard one. Image quality was greatly improved using prospective motion-correction in the case of deliberate motion during the scan. |
| 1710 | Computer 79
|
More Efficient SPGR 3D Abdominal Imaging Using an Interleaved Randomized Spoiler |
| zheng zhong1, Christopher M. Sandino2, Ali Syed1, John M. Pauly2, and Shreyas S. Vasanawala1 | ||
1Radiology, Stanford University, Stanford, CA, United States, 2Electrical Engineering, Stanford University, Stanford, CA, United States |
||
3D SPGR is commonly used clinically for contrast-enhanced body imaging, but requires time for a gradient spoiler every TR. This consumes up to 40% of scan time, even for a non-cartesian trajectories with relatively long readouts, such as cones. Removing the spoiler can save time but causes severe artifacts. In this study, we propose a novel approach of applying interleaved, randomized spoilers to a motion-robust 3D-cones sequence. The feasibility of the approach to shorten acquisition time without introducing image artifacts was validated on both healthy subjects and patients. |
| 1711 | Computer 80
|
Optimal Control of the contrast in a MP-RAGE sequence: an application on the pelvis |
| Benoît Vernier1,2, Eric Van Reeth1,3, Marc Lapert2, Olivier Hamelin4, Olivier Beuf1, Hélène Ratiney1, and Frank Pilleul4 | ||
1CREATIS, Lyon, France, 2Siemens Healthcare SAS, Saint-Denis, France, 3CPE, Lyon, France, 4Centre Léon Bérard, Lyon, France |
||
MRI contrast enhancement by Optimal Control is a new approach to design optimal magnetization preparation. This allows to maximize the contrast between target tissues characterized by their relaxation times. Recent numerical implementations have made possible the optimization of such preparation in a steady state sequence without full recovery between each repetition. This abstract demonstrates the contrast flexibility offered by this approach when combined with a MPRAGE sequence on pelvis imaging at 3T. |
| 1712 | Computer 81
|
Tennisball: slew-efficient trajectory design for 3D-radial imaging |
| Gopal Nataraj1 and Michael Lustig1 | ||
1University of California, Berkeley, Berkeley, CA, United States |
||
What is the most efficient path to traverse a sphere? For a spherical gap-minimizing efficiency definition and a particular path length, the answer happens to resemble the seam lines of a tennis ball. Here we explore properties of this general class of spherical trajectories and investigate their application to slew-constrained ZTE imaging. Static and motion-resolved imaging results demonstrate that Tennisball trajectories can achieve competitive performance with existing methods but with ~30-40% lower peak slew. |
| 1713 | Computer 82
|
Fast 3D Isotropic High-Resolution MRI of Mouse Brain Using a Variable Flip Angle RARE Sequence With T2 Compensation @9.4T |
| Lisa Y Hung1, Karl-Heinz Herrman1, and Jürgen R Reichenbach1 | ||
1Medical Physics Group, Institute of Diagnostic Radiology, Jena University Hospital, Jena, Germany |
||
Some applications require T2-weighted mouse brain MRI with isotropic high resolution (100µm3), but at 9.4T the short T2 of mouse brain limits the echo train length of RARE sequences and exacerbates T2-decay artifacts. Adapting a RARE sequence by using variable flip-angles allows a reduction in scan time from 19min to 7-9min with similar image quality. However, long RARE echo trains and strong imaging gradients can unintentionally cause substantial diffusion weighting. To minimize the diffusion effects crusher gradients were optimized and the variable flip angles were kept above 40°. |
| 1714 | Computer 83
|
Simultaneous T2-weighted Real-time MRI of 2 Orthogonal Slices |
| Samantha Hickey1, Andreas Reichert1, Thomas Bortfeld2, and Michael Bock1 | ||
1Dept. of Radiology, Medical Physics, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 2Division of Radiation Biophysics, Department of Radiation Oncology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States |
||
In cancer radiotherapy, MR-guidance allows for superior tumor visualization and real-time target volume tracking. Real-time pulse sequences for MR-guidance require high acquisition speeds and good tumor-to-background contrast, which is often achieved with T2-weighted acquisitions. Here, we combine a real-time sequence with orthogonal slices with a T2w echo-shifted contrast and demonstrate its use under breathhold and free-breathing conditions in a volunteer. |
| 1715 | Computer 84
|
Bipolar Partial Echo Imaging |
| Holger Eggers1 and Jochen Keupp1 | ||
1Philips Research, Hamburg, Germany |
||
Partial echo imaging is widely used in angiography to shorten the echo time and reduce flow-induced signal loss. However, sampling only an asymmetric part of k-space basically ignores high-frequency phase and thus causes signal alteration as well. In the present work, bipolar partial echo imaging is introduced to cover the full k-space sparsely with a Cartesian partial echo measurement. It involves reversing the polarity of the readout gradient during the acquisition and recovering missing data in the reconstruction without assuming smoothness of the phase. The potential of bipolar partial echo imaging is demonstrated in phantom and first volunteer experiments. |
| 1716 | Computer 85
|
Susceptibility weighted imaging using Stack of Spiral trajectories |
| Tzu-Cheng Chao1 and James G. Pipe1 | ||
1Department of Radiology, Mayo Clinic, Rochester, MN, United States |
||
A new strategy to acquire Susceptibility Weighted Imaging was proposed with Stack of Spiral trajectories to reduce scan time while maintaining adequate SNR. Spiral encoding offers acceleration in scan speed by six times compared to conventional Cartesian methods, with similar SNR, resolution, and image quality. Due to its improved SNR efficiency, the proposed method has the potential to facilitate lesion detection with higher resolution SWI in a reasonable scan time. |
| 1779 | Computer 49
|
Attenuation Correction in PET/MRI Pediatric Treated Brain Tumors: a Preliminary Comparison between the ZTE and Atlas Techniques |
| Paola Scifo1, Federico Fallanca1, Maurizio Barbera2, Annarita Savi1, Valentino Bettinardi1, Ilaria Neri1, Giovanna Gattuso3, Maura Massimino3, Paolo Verderio4, Sara Pizzamiglio4, Andrea Falini2,5, Luigi Gianolli1, Filippo Spreafico3, Maria Picchio1,5, and Cristina Baldoli2,5 | ||
1Nuclear Medicine, IRCCS San Raffaele Scientific Institute, Milan, Italy, 2Neuroradiology Unit, IRCCS San Raffaele Scientific Institute, Milan, Italy, 3Department of Medical Oncology and Hematology, Pediatric Oncology Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy, 4Unit of Bioinformatics and Biostatistics, Department of Applied Research and Technological Development, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy, 5Vita-Salute San Raffaele University, Milan, Italy |
||
A preliminary comparison between the two Atlas and ZTE attenuation correction methodologies in a group of pediatric patients with brain tumors is presented. The differences of the two ACmaps and PET images were calculated and the PET parameters were compared. Our results show that using ZTE or Atlas to generate ACmaps modifies the SUVs, but the differences in values are very small. Moreover, when metallic artefacts occur, both ZTE and Atlas show loss of signals but ACmapsAtlas has smaller artefacts. |
| 1780 | Computer 50
|
Evaluating b-value combinations on IVIM-DKI analysis with parametric reconstruction method in pancreatic carcinoma: A digital phantom study |
| Archana Vadiraj Malagi1, Devasenathipathy Kandasamy2, Kedar Khare3, Fernando Calamante4, Raju Sharma2, and Amit Mehndiratta5,6 | ||
1Centre for Biomedical Engineering, Indian Institute of Technology Delhi, Navi Mumbai, India, 2Department of Radiodiagnosis, All India Institute of Medical Sciences Delhi, New Delhi, India, 3Department of Physics, Indian Institute of Technology Delhi, New Delhi, India, 4Sydney Imaging and School of Biomedical Engineering, University of Sydney, Sydney, Australia, 5Indian Institute of Technology Delhi, New Delhi, India, 6Department of Biomedical Engineering, All India Institute of Medical Sciences Delhi, New Delhi, India |
||
Optimization of b-values in IVIM-DKI is required for parameter estimation accuracy. IVIM-DKI model with Total Variation correction approach was employed in this study, which avoids abrupt changes in parameter values and eliminates non-physiological heterogeneity in maps. In pancreatic cancer, different b-value combinations were used: 4,6,8,10, and 13b-values. In terms of accurate parameter estimations and comparable qualitative maps, 8-13 b-values exhibited a similar trend; in contrast, 4b-values indicated inaccurate and noisy maps. 8b-values can be employed in pancreatic cancer with a constrained technique for a shorter acquisition process and reduced noise in parameter maps at low SNR. |
| 1781 | Computer 51
|
A Novel Acquisition Strategy for Volumetric Diffusion-Weighted Imaging using Variable Density TSE-CASPR |
| Sebastian Eduardo Fonseca1, Sheng-Qing Lin1, Limin Zhou1, Yiming Wang1, and Ananth J Madhuranthakam1,2 | ||
1Department of Radiology, UT Southwestern Medical Center, Dallas, TX, United States, 2Advanced Imaging Research Center, UT Southwestern Medical Center, Dallas, TX, United States |
||
Diffusion weighted imaging (DWI) is a highly valuable MR imaging technique in the clinic. Single-shot EPI is the reference standard sequence for DWI due to its speed, however it suffers from severe distortion artifacts that can obscure potential pathologies. In this abstract, we present a novel 3D acquisition strategy that uses a fast spin echo readout and variable density sampling for volumetric DWI. The oversampled center region serves as navigator to correct phase errors. We present preliminary results in brains of healthy volunteers compared to the standard sequences used in DWI. |
| 1782 | Computer 52
|
Self-Gated Zero Echo Time Imaging of the Lung |
| Hanna Frantz1, Tobias Lobmeyer1, Patrick Metze1, Thomas Hüfken1, Kilian Stumpf1, Julien Rivoire2, Hizami Murad2, and Volker Rasche1 | ||
1Department of Internal Medicine II, Ulm University Medical Center, Ulm, Germany, 2RS2D, Mundolsheim, France |
||
Zero echo time (ZTE) imaging is an established approach for three-dimensional imaging of tissues and materials with ultrashort T2* relaxation times. An intrinsic disadvantage of this approach are missing data points in the center region of k-space due to hardware limitations. These missing points are often retrospectively interpolated or subsequently acquired using single-point imaging approaches or additional ZTE acquisitions with decreased resolutions. The presented approach relies on an interleaved combination of ZTE read-outs with different resolutions in combination with Compressed Sensing, to acquire the k-space center, thus enabling k-space based self-gating. |
| 1783 | Computer 53
|
Ultra-Short Echo Time 31P 3D MRSI at 3T with Novel Rosette k-space Trajectory |
| Brian Matthew Bozymski1, Xin Shen2, Ali Caglar Özen3, Serhat Ilbey3, Albert Thomas4, Mark Chiew5, Ulrike Dydak1,6, and Uzay Emir1,2 | ||
1School of Health Sciences, Purdue University, West Lafayette, IN, United States, 2Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States, 3Department of Radiology, Medical Center, University of Freiburg, Freiburg, Germany, 4Department of Radiology, University of California, Los Angeles, CA, United States, 5Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom, 6Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN, United States |
||
Feasibility of a novel, rosette trajectory UTE (70 μs) 3D 31P MRSI sequence is tested at 3T in phantoms and in upper leg scan with healthy subject. Braino phantom, proton resolution phantom, and highly concentrated 31P phantom test demonstrated proper reconstruction and metabolite localization. In vivo calculations of acquired PCr and β-ATP signals showed competitive SNR across 20 voxels of interest. With further acceleration, the sequence may serve as a superior alternative to conventional weighted 3D MRSI, allowing for greater SNR or resolution in briefer times. Notably, this unprecedentedly short UTE 31P MRSI intrinsically avoids baseline and phasing issues. |
| 1784 | Computer 54
|
Insights to the diffusion MRI ‘crossing fiber problem’: Characterizing the micro-structure in an MRI voxel using synchrotron radiation imaging |
| Hans Martin Kjer1,2, Mariam Andersson2, Alexandra Pacureanu3, Vedrana Andersen Dahl1, Anders Bjorholm Dahl1, and Tim Bjørn Dyrby1,2 | ||
1Technical University of Denmark, Kgs. Lyngby, Denmark, 2Danish Research Centre for Magnetic Resonance, Hvidovre, Denmark, 3European Synchrotron and Radiation Facility, Grenoble, France |
||
Complex fiber regions where multiple tracts and bundles intersect are challenging to study with diffusion weighted MRI (DWI), due to the anatomical variation and 3D complexity within the volume covered by the measure voxel. Using x-ray nano-holotomography (XNH) we have been able to obtain and analyze a volume from a complex fiber region and make a comparison against DWI measurements from the same location of the same ex-vivo monkey brain. We derive comparable micro-structural features in the form of orientation distributions and anisotropy measures and show strong connections between them. |
| 1785 | Computer 55
|
Benchmarking learned non-Cartesian k-space trajectories and reconstruction networks |
| Chaithya G R1,2 and Philippe Ciuciu1,2 | ||
1NeuroSpin, Joliot, CEA, Université Paris-Saclay, F-91191, Gif-sur-Yvette, France, 2Inria, Parietal, Université Paris-Saclay, F-91120, Palaiseau, France |
||
We benchmark the current existing methods to jointly learn non-Cartesian k-space trajectory and reconstruction: PILOT1, BJORK2 and compare them with those obtained from recently developed generalized hybrid learning (HybLearn) framework3. We present the advantages of using projected gradient descent to enforce MR scanner hardware constraints as compared to using added penalties in the cost function. Further, we use the novel HybLearn scheme to jointly learn and compare our results through retrospective study on fastMRI validation dataset. |
| 1786 | Computer 56
|
Evaluation of Multicoil SENSE Reconstruction of Undersampled Data using a Human Observer Model of Signal Detection |
| Alexandra G O'Neill1, Tavianne M Kemp1, Sajan G Lingala2, and Angel R Pineda1 | ||
1Mathematics Department, Manhattan College, Riverdale, NY, United States, 2Roy J. Carver Department of Biomedical Engineering, University of Iowa, Iowa City, IA, United States |
||
We evaluated undersampling in MRI using a multicoil SENSE reconstruction with no regularization based the detection of signals by humans. We used a sparse difference-of-Gaussians (S-DOG) model to predict human performance in the detection of a small and large signal in anatomical backgrounds. The prediction was then validated using human observer two-alternative forced choice (2-AFC) tasks. Our model predicted a decrease in performance for both the small and large signal from 4X to 5X acceleration. Our observer study validated that prediction. This approach may lead to a way of assessing image quality that predicts human performance with fewer observer studies. |
| 1787 | Computer 57
|
Optimal Phase-Encoding Schemes for Regional Bias and Reproducibility of Diffusion MRI derived Measures |
| M. Okan Irfanoglu1, Grayson Clark1, and Carlo Pierpaoli1 | ||
1QMI, NIBIB/NIH, Bethesda, MD, United States |
||
Our goal is to determine whether diffusion MRI acquisitions with different phase-encoding directions (PEDs) result in comparable quantitative measures in different regions of the human brain, or whether PEDs introduce a bias. Additionally, we aim to determine if for a given region, a set of PEDs has better reproducibility. For several regions, PA phase-encoding direction resulted in values that were significantly different than the other three. LR/RL phase-encoding was more reproducible compared to AP/PA in most brain regions, with frontal white matter being the exception. The main source of these reproducibility variations was identified as ghosts overlapping with brain regions. |
| 1788 | Computer 58
|
Predicting artifacts in maximum intensity projections of high b-value DWI of the breast using neural networks. |
| Andrzej Liebert1, Lorenz Kapsner1, Lukas Folle2, Hannes Schreiter1,2, Badhan Kumar Das1,2, Sabine Ohlmeyer1, Andreas Maier2, Evelyn Wenkel1, Frederik B. Laun1, Michael Uder1, and Sebastian Bickelhaupt1 | ||
1Institute of Radiology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany, 2Pattern Recognition Lab, Department of Computer Science, Friedrich-Alexander University Erlangen Nuremberg, Erlangen, Germany |
||
DWI acquisitions in MRI are prone to artifacts caused e.g. by insufficient fat saturation and can significantly impede the diagnostic assessment. We investigated the ability to predict the occurrence of these artifacts in maximum intensity projections (MIPs) of high b-value DWI using a neuronal network that analyses T2w images, which were acquired prior to the DWI sequence. An AuRoC of 0.83 with sensitivity of 0.82 and specificity of 0.61 was achieved. Investigation of the regions of the T2w images most important for the decision of the neural network showed a good correlation with artifact-affected areas in the DWI MIP images. |
| 1789 | Computer 59
|
Predicting Perfusion Augmentation Using Deep Learning without Vasodilators |
| Moss Zhao1, Ramy Hussein1, Michael Moseley1, and Greg Zaharchuk1 | ||
1Department of Radiology, Stanford University, Stanford, CA, United States |
||
We present a deep learning technique to predict cerebral perfusion after vasodilation challenges. A 3D convolutional neural network (CNN)-based encoder-decoder architecture was constructed to transform ASL perfusion images acquired pre-vasodilation into post-vasodilation images using an improved attention-gated 3D U-Net. Results showed that the prediction and ground truth were not significantly different. This technique will enable a drug-free MR procedure to study the hemodynamic of patients with high risk cerebrovascular diseases. |
| 1790 | Computer 60
|
Convolutional Neuronal Network Inception-v3 detects Partial Volume Artifacts on 2D MR-Images of the Lung for Automated Quality Control |
| Andreas Voskrebenzev1,2, Cristian Crisosto1,2, Maximilian Zubke1,2, Frank Wacker1,2, and Jens Vogel-Claussen1,2 | ||
1Institute of Diagnostic and Interventional Radiology, Hannover Medical School, Hannover, Germany, 2German Center for Lung Research (DZL), Biomedical Research in Endstage and Obstructive Lung Disease Hannover (BREATH), Hannover, Germany |
||
The partial volume effect (PVE) is an often-observed artifact in MR imaging. Especially images with a low spatial resolution, will show an averaged voxel signal of multiple tissue components. These artifacts can be so substantial that a further image analysis can be omitted. This is e.g. the case for phase-resolved functional lung imaging (PREFUL), which is based on the 2D acquisition of coronal image-time-series to assess ventilation and perfusion dynamics. In this study the pretrained convolutional neural network Inception-v3 was trained via transfer-learning to detect images, which show substantial PVE with a classification accuracy of 91%. |
| 1791 | Computer 61
|
Multi-Scale Evaluation of Uncertainty Quantification Techniques for Deep Learning based MRI Segmentation |
| Benjamin Lambert1,2, Florence Forbes3, Alan Tucholka2, Senan Doyle2, and Michel Dojat1 | ||
1Univ. Grenoble Alpes, Inserm, U1216, Grenoble Institut Neurosciences, GIN, 38000, Grenoble, France, 2Pixyl, Research and Development Laboratory, 38000, Grenoble, France, 3Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK, 38000, Grenoble, France |
||
Efforts are required to design Deep Learning models that are not only powerful, but also capable of expressing the certainty of their predictions. We evaluate 3 state-of-the-art techniques for uncertainty quantification : Monte-Carlo Dropout, Deep Ensemble, and Heteroscedastic models. Evaluation is illustrated on a task of automatic segmentation of White-Matter Hyperintensities in T2-weighted FLAIR MRI sequences of Multiple-Sclerosis patients. Analysis is performed at 3 different scales : the voxel, the lesion, and the whole image. Results indicate the superiority of the Heteroscedastic approach, which ranked first in both the uncertainty and segmentation tasks. |
| 1792 | Computer 62
|
Grey Matter Connectome Data Analysis and Classification of Multiple Sclerosis Clinical Profiles |
| Berardino Barile1, Claudio Stamile2, Sabine Van Huffel3, and Dominique Sappey-Marinier4 | ||
1CREATIS (UMR5220 CNRS & U1294 INSERM, Université Claude Bernard-Lyon1 & INSA-Lyon, France, Villeurbanne, France, 2R&D Department CGnal Milan, Italy, Villeurbanne, France, 3KU Leuven, Leuven, Belgium, 4CREATIS (UMR5220 CNRS & U1294 INSERM), Université Claude Bernard-Lyon1 & INSA-Lyon, France, Villeurbanne, France |
||
Multiple Sclerosis (MS) is an autoimmune inflammatory disease characterized by white matter (WM) lesions and gray matter (GM) degeneration leading to cognitive and physical impairments. Recently, brain connectivity techniques were developed to better characterize such brain damages in association with the clinical profiles of MS patients. In this study, we proposed a complete automated pipeline to first extract global graph metrics from the morphological brain connectivity based on GM tissue thickness and second to train a higher-level ensemble of four machine learning models for the classification of MS clinical profiles using only the T1-weighted image modality. |
| 1793 | Computer 63
|
Analysing The Role Of Model Uncertainty in Flourine-19 MRI using Markov Chain Monte Carlo methods |
| Masoumeh Javanbakhat1, Ludger Starke2, Sonia Waiczies2, and Christoph Lippert1 | ||
1Digital Health-Machine Learning Group, Hasso Plattner Institute, Potsdam, Germany, 2Berlin Ultrahigh Field Facility, Max Delbrück Centre for Moleculare Medicine in the Helmholtz Association, Berlin, Germany |
||
Deep learning (DL) has achieved state of the art results in semantic segmentation of numerous medical imaging applications. Despite promising results deep learning models tend to produce point estimates as outputs which leads to overconfident, miscalibrated predictions. These overconfident predictions are specifically problematic in medical applications. Hence, providing a measure of a system’s confidence to identify untrustworthy predictions is essential to guide clinical decisions. Here we propose a 3D Bayesian segmentation model to provide uncertainty estimation for the Fluorine-19 MRI dataset based on Stochastic Gradient Markov Chain Monte Carlo methods. |
| 1794 | Computer 64
|
Resting-state functional connectivity predicts subsequent pain-related threat learning |
| Balint Kincses1,2, Katarina Forkmann1, Katharina Schmidt1, Ulrike Bingel1, and Tamas Spisak2 | ||
1Bingel-laboratory, Department of Neurology, University Hospital Essen, Essen, Germany, 2Laboratory of Predictive NeuroImaging, Institute for Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany |
||
Fear conditioning has a role in anxiety-disorders and the neurobiological correlates of it are not yet well understood. Therefore, we trained a machine learning predictive model on individual functional resting state connectivity data to predict the emotional aspects of fear conditioning. The model was found to predict individual pain-related threat learning measured by the change of valence with an explained variance of 24%-41%. These results highlight the potential of machine learning to enhance our understanding of fear conditioning. |
| 1795 | Computer 65
|
Developing pattern recognition models to extract longitudinal network-based measures at an individual level |
| Elisa Colato1, Claudia AM Wheeler-Kingshott1,2,3, Douglas L Arnold4, Frederik Barkhof1,5,6,7, Olga Ciccarelli1,5, Declan Chard1,5, and Arman Eshaghi1,8 | ||
1Queen Square Multiple Sclerosis Centre, Department of Neuroinflammation, Queen Square Institute of Neurology, Faculty of Brain Sciences, University College London (UCL), London, United Kingdom, 2Brain MRI 3T Research Centre, C. Mondino National Neurological Institute, Pavia, Italy, 3Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, 4McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, QC, Canada, 5Institute for Health Research (NIHR), University College London Hospitals (UCLH) Biomedical Research Centre (BRC), London, United Kingdom, 6Department of Radiology and Nuclear Medici, VU medical centre, Amsterdam, Netherlands, 7Centre for Medical Image Computing (CMIC), Department of Medical Physics and Biomedical Engineering, University College London (UCL), London, United Kingdom, 8Centre for Medical Image Computing (CMIC), Department of Computer Science, University College London (UCL), London, United Kingdom |
||
Network-based measures can outperform regional and whole-brain grey matter (GM) measures in explaining clinical disability in several neurodegenerative disorders. However, network measures are mostly estimated at the group level and require a re-estimation of model parameters when applied to new participants. Here, we introduce a new longitudinal network analysis paradigm to extract longitudinal ICA-like components at an individual level from a discovery cohort, applied machine learning to obtain individual-level network-based measure for a validation cohort, and used them to explain clinical disability in multiple sclerosis. |
| 1796 | Computer 66
|
Deeping-Learning based acceleration for Serial MR imaging using Transformer (SMR-Transformer): Comparison with CNN on Three Datasets |
| Chuyu Liu1, Shuo Chen1, Yibing Chen2, Rui Li1, and Xiaolei Song1 | ||
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, Tsinghua University, Beijing, China, 2Xi’an Key Lab of Radiomics and Intelligent Perception, School of Information Sciences and Technology, Northwest University, Xi'an, China |
||
Serial image acquisition is always required for quantitative and dynamic MRI, in which the spatial and temporal resolution as well as the contrast is restricted by acquisition speed. Herein we employed a novel transformer based deep-learning method for reconstruction of highly under-sampled MR serial images. The network architecture and input were designed to take advantages of Transformer’s capacity on global information learning along the redundant series dimension. SMR-transformer was compared with 3D Res-U-Net, on a self-acquired CEST dataset, a cardiac dataset from OCMR and a diffusion dataset from HCP. Both images and quantitative comparison indicated SMR-Transformer allowed fast serial MRI. |
| 1858 | Computer 38
|
Development of Non-Rectangular Excitation Pulses for Improved kZ Fidelity in Super-Resolution T2-Weighted Spin-Echo MRI |
| Eric Allen Borisch1, Alexandru C. Florea1,2, Roger C. Grimm1, Akira Kawashima3, and Stephen J. Riederer1 | ||
1Radiology, Mayo Clinic, Rochester, MN, United States, 2Gustavus Adolphus College, Saint Peter, MN, United States, 3Radiology, Mayo Clinic, Phoenix, AZ, United States |
||
Results from continuing investigation of non-rectangular slice excitation profiles with the goal of providing improved results when used to reconstruct overlapping-slice 2D acquisitions leveraging knowledge of the slice profile are presented. A simulation of the excitation/acquisition process and accompanying reconstruction has been developed to aid the exploration of non-traditional slice profiles and their impact on the final reconstruction quality. |
| 1859 | Computer 39
|
Accelerated Magnetic Resonance Imaging with Flow-Based Priors |
| Frederik Fraaz1 and Reinhard Heckel1,2 | ||
1Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany, 2Department of Electrical and Computer Engineering, Rice University, Houston, TX, United States |
||
Convolutional neural networks trained end-to-end achieve state-of-the-art image quality for accelerated MRI. But end-to-end networks are trained for a specific undersampling operator. A more flexible approach that can work with any undersampling operator at inference is to train a generative image prior and impose it during reconstruction. In this work, we train a flow-based generator on image patches and then impose it as a prior in the reconstruction. We find that this method achieves slightly better reconstruction quality than state-of-the-art un-trained methods and slightly worse quality than neural networks trained end-to-end on the 4x accelerated multi-coil fastMRI dataset. |
| 1860 | Computer 40
|
Radial Echo-Planar Imaging with Subspace Reconstruction for Brain MRI |
| Zhengguo Tan1 and Martin Uecker1,2 | ||
1Institute for Diagnostic and Interventional Radiology, University Medical Center Goettingen, Goettingen, Germany, 2Institute of Medical Engineering, Graz University of Technology, Graz, Austria |
||
While echo planar imaging (EPI) is successful in many MRI applications, this study demonstrates that radial EPI is applicable to brain MRI as well. Initial volumetric brain experiments at 1 mm isotropic resolution show that radial EPI is capable of rendering high-quality T1- and T2*-weighted images at 2.5 and 1.3 minutes scan time, respectively. Further, radial EPI requires neither magnetization preparation nor fat saturation. |
| 1861 | Computer 41
|
Robust and fast RF pulse design of rectangle excitations using a hybrid CNN framework with signal modeling |
| Yajing Zhang1, Guohui Ruan1, Yan Zhao2, Xianrui Huang2, and Marc Van Cauteren3 | ||
1BU-MR Clinical Science, Philips Healthcare, Suzhou, China, 2BU-MR R&D, Philips Healthcare, Suzhou, China, 3BU-MR Clinical Science, Philips Healthcare, Tokyo, Japan |
||
Radio frequency (RF) pulses play a key role in magnetic resonance imaging. However, current MRI systems mostly use pre-defined RF waveforms with minimal adaptions. In this work, we propose a novel hybrid framework to combine convolutional neural network (CNN) with signal modeling by feature extractions. The result demonstrates that the proposed method provides more robust pulse design estimation result than using a generalized CNN framework for any desired rectangle excitations. |
| 1862 | Computer 42
|
Practical RF-pulse shape designs to minimize off-resonance artifacts in dissolved-phase hyperpolarized 129Xe MRI |
| Junlan Lu1, Suphachart Leewiwatwong2, Bastiaan Driehuys2, and John Mugler3 | ||
1Medical Physics, Duke University, Durham, NC, United States, 2Biomedical Engineering, Duke University, Durham, NC, United States, 3Radiology and Medical Imaging, University of Virginia, Charlottesville, VA, United States |
||
Minimizing artifacts in 129Xe gas exchange MRI requires selectively exciting the broad dissolved phase resonances, without exciting the much larger gas phase pool. This requires not only a short, selective pulse, but also overcoming pulse distortions imposed by the RF amplifier. Here, we demonstrate a novel, practical RF pulse based on augmenting the traditional Hanning window pulse with tunable DC bias and sub-lobes in the time-domain. We characterized the performance through simulations, phantom, and in vivo testing and determined that the tunable DC bias parameter creates a robust dissolved-phase excitation while minimizing gas-phase contamination. |
| 1863 | Computer 43
|
Reduction of Acquisition time using RAPID-SI technique: Preliminary in-vitro results and comparison with CSI at Ultra-High-Field. |
| Hacene Serrai1 | ||
1Carle Foundation Hospital, Center for Clinical Imaging Research, Urbana, IL, United States |
||
It has been shown that RAPID-SI, implemented at 3 Tesla magnet, allows for reduction of acquisition time while providing accurate metabolite information as compared to CSI. To benefit from the higher sensitivity of the Ultra-High-Field (UHF), RAPID-SI was implemented on a Siemens Terra 7T scanner, and phantom MRSI data were acquired, with less averaging, and compared in terms of acquisition time, signal-to-noise ratio (SNR), and data analysis to those collected using CSI method. As expected, and in addition to the acquisition time reduction, proportional to the speed factor R, RAPID-SI provides similar quantification results with similar SNR values, afterward data reconstruction, as compared to CSI. |
| 1864 | Computer 44
|
Characterization of SNR for radial spin echo imaging |
| Neta Stern1, Kai T Block2, Chen Solomon1, Tamar Blumenfeld-Katzir1, and Noam Ben-Eliezer1,2,3 | ||
1Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel, 2New-York University Langone Medical Center, New York, NY, United States, 3Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel |
||
To optimize the use of radial spin-echo sequences, SNR characterization for relevant T1 and T2 ranges is required. In this work, we studied the interplay between the number of excitations (Nexc) and the turbo factor (NTF - number of echoes within each TR) on the SNR of images acquired using the RAISE (RAdIal Spin Echo) sequence. SNR was evaluated for phantom and in-vivo images. Phantom analysis focused on SNR per T1 and T2 ranges appropriate for known gray and white matter relaxation times. Selected parameter sets were later used for in vivo scans for qualitative evaluation. |
| 1865 | Computer 45
|
Single-Shot Multi-b-value (SSMb) Diffusion-weighted MRI Using Spin Echo and Stimulated Echoes with Variable Flip Angles |
| Guangyu Dan1,2, Kaibao Sun1, Qingfei Luo1, and Xiaohong Joe Zhou1,2,3 | ||
1Center for Magnetic Resonance Research, University of Illinois at Chicago, Chicago, IL, United States, 2Department of Biomedical Engineering, University of Illinois at Chicago, Chicago, IL, United States, 3Departments of Radiology and Neurosurgery, University of Illinois at Chicago, Chicago, IL, United States |
||
In diffusion-weighted imaging with multiple b-values, the acquisition times using a spin-echo EPI sequence can be long because each b-value requires a new shot. The lengthy acquisition time imposes issues with patient compliance and increases venerability to motion-induced errors in quantitative analysis. We herein report a novel pulse sequence that acquires multiple diffusion-weighted images with difference b-values in a single shot by utilizing the combination of a spin echo and a train of stimulated echoes with variable flip angles. This sequence has been demonstrated in a quantitative diffusion phantom and the human brain to realize a 4-fold scan time reduction. |
| 1866 | Computer 46
|
Diffusivity-limited q-space trajectory imaging |
| Deneb Boito1,2, Magnus Herberthson3, and Evren Özarslan1,2 | ||
1Department of Biomedical Engineering, Linköping University, Linköping, Sweden, 2Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, Sweden, 3Department of Mathematics, Linköping University, Linköping, Sweden |
||
Q-space trajectory imaging (QTI) characterizes microstructures through the statistical moments of the diffusion tensor distribution. A constrained estimation framework named QTI+ was recently proposed to achieve mathematically and physically acceptable estimates of these moments. Here we consider expanding QTI+ with a new set of conditions based on the theoretical maximum value of water diffusivity. We show where these conditions are violated for different bulk diffusivity values, and how the new constraints affect the QTI scalar maps. The results show violations occurring almost exclusively in voxels containing gray matter and CSF. Imposing the constraints produces metrics closer to the expected values. |
| 1867 | Computer 47
|
Quantitative Reduced Field-of-view Imaging using 3D Tailored Inner Volume Excitation and Pattern Recognition |
| Yun Jiang1, Jon-Fredrik Nielsen2, Vikas Gulani1, and Nicole Seiberlich1 | ||
1Department of Radiology, University of Michigan, Ann Arbor, MI, United States, 22Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States |
||
The goal of this study is to achieve small field-of-view (FOV) T1 and T2 mapping using 3D tailored inner volume (IV) excitation and pattern recognition based on MR Fingerprinting. |
| 1868 | Computer 48
|
Pressure Pain Activation Revealed in Simultaneous Brain, High-Res Brainstem, and High-Res Spinal Cord fMRI. |
| Christine S W Law1, Ken A Weber1, Sean Mackey1, and Gary H Glover1 | ||
1Stanford University, Stanford, CA, United States |
||
We present a novel fMRI acquisition that mitigates susceptibility-induced off-resonance and simultaneously captures brainstem, spinal cord, and brain. Brainstem and spinal cord are imaged in high spatial resolution whereas brain resolution is standard. This fMRI acquisition is applied in a noxious pressure-pain experiment from which we observe activation in brain, spinal cord, and brainstem pons & medulla. |
| 1869 | Computer 49
|
Comparison of Imaging Methods for Fast Whole-Body MRI Screening |
| Marc Rea1, Alex Barrett1, and Christopher Romaniuk2 | ||
1Physics, Clatterbridge Cancer Centre, Liverpool, United Kingdom, 2Radiology, Clatterbridge Cancer Centre, Liverpool, United Kingdom |
||
Improvements in image quality and scan time coupled with AI reading of images, is enabling the possibility of quickly screening all patients who attend MRI for malignant tumours. This research explores the imaging and clinical requirements for such a programme, and reports on the ability of various sequences and patient configurations to meet those requirements. |
| 1870 | Computer 50
|
SSFD: Self-Supervised Feature Distance Outperforms Conventional MR Image Reconstruction Quality Metrics |
| Philip M. Adamson1, Jeffrey Dominic1, Arjun Desai1, Christian Bluethgen2, Jeff P. Wood3, Ali B. Syed2, Robert Boutin2, Kathryn J. Stevens2, Daniel Spielman2, Shreyas Vasanawala2, John M. Pauly1, Akshay S. Chaudhari2, and Beliz Gunel1 | ||
1Department of Electrical Engineering, Stanford University, Palo Alto, CA, United States, 2Department of Radiology, Stanford University, Palo Alto, CA, United States, 3Austin Radiological Association, Austin, TX, United States |
||
Evaluation of accelerated magnetic resonance imaging (MRI) reconstruction methods is imperfect due to the discordance between quantitative image quality metrics (IQMs) and radiologist-perceived image quality. Self-supervised learning (SSL) is a deep learning (DL) method that has become a popular pre-training tool due to its ability to capture generalizable and domain-specific feature representations of the underlying data without the need for labels. In this study, we derive a data-driven self-supervised feature distance (SSFD) IQM to assess MR image reconstruction quality. We demonstrate that SSFD is more highly correlated to three radiologist’s perceived image quality on DL-based sparse reconstructions than conventional IQMs. |
| 1871 | Computer 51
|
Impact Of MR Intensity Normalization For Different MR Sequences In MRI Based Radiomics Studies |
| Patrick Salome1,2,3,4, Francesco Sforazzini1,2,3,4, Andreas Kudak3,5,6, Matthias Dostal3,5,6, Nina Bougatf3,5,6, Jürgen Debus3,4,5,7, Maximilian Knoll1,3,4,5, and Amir Abdollahi1,3,4,5 | ||
1CCU Translational Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany, 2Medical Faculty, Heidelberg University Hospital, Heidelberg, Germany, 3Heidelberg Ion-Beam Therapy Center (HIT), Heidelberg, Germany, 4German Cancer Consortium (DKTK) Core Center, Heidelberg, Germany, 5Radiation Oncology, Heidelberg University Hospital, Heidelberg, Germany, 6CCU Radiation Therapy, German Cancer Research Center (DKFZ), Heidelberg, Germany, 7National Center for Tumor Diseases (NCT), Heidelberg, Germany |
||
MR-based radiomics prognostic signatures have great clinical potential but are currently hindered due to a lack of standardisation in the radiomics workflow. This study focuses on a crucial step of these workflows, i.e. intensity normalisation, while presenting a methodology that allows for determining the most suitable MR intensity normalization method for a specific entity. |
| 1872 | Computer 52
|
Cortical thickness predicts pain sensitivity in a multi-centre cohort: a machine learning approach |
| Raviteja Kotikalapudi1,2, Balint Kincses1,2,3, Kevin Hoffschlag1, Matthias Zunhammer2, Tobias Schmidt-Wilcke4,5, Zsigmond T Kincses3, Ulrike Bingel2, and Tamas Spisak1,2 | ||
1Laboratory of Predictive NeuroImaging, Institute for Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany, 2The Bingel Laboratory, Translational Pain Research Unit, University Hospital Essen, Essen, Germany, 3Department of Neurology, University of Szeged, Szeged, Hungary, 4Institute for Clinical Neurosciences and Medical Psychology, Heinrich Heine University, Dusseldorf, Germany, 5Neurocentre, District Hospital Mainkofen, Mainkofen, Deggendorf, Germany |
||
Individual sensitivity to pain is both a precursor and a symptom of many clinical pain conditions. A pain predictive model would have potential applications in objectively characterizing pain in acute and chronic pain individuals. Here, we developed a cortical thickness-based predictive model of pain sensitivity using a machine learning approach and multi-centre T1-weighted MRI and quantitative pain threshold data. We found that our model significantly predicts pain sensitivity, that was measured through heat, cold and mechanical stimuli. Furthermore, the predictions were exclusively driven by cortical thickness and not confounded by variables of demographic and psychological value. |
| 1873 | Computer 53
|
Benchmarking Accelerated MRI: A Head-to-Head Comparison of Deep Learning Reconstruction and Super-Resolution Techniques |
| Eric K. Gibbons1, Zhongnan Fang2, Arjun D. Desai3, Christopher M. Sandino3, Garry E. Gold4, Brian A. Hargreaves4, and Akshay S. Chaudhari4 | ||
1Department of Electrical and Computer Engineering, Weber State University, Ogden, UT, United States, 2Lvis Corporation, Palo Alto, CA, United States, 3Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 4Department of Radiology, Stanford University, Stanford, CA, United States |
||
Deep-learning (DL) can be used to extend compressed sensing (CS) to learn the regularization function in a data-driven manner. In contrast, super resolution (SR) algorithms have been used to transform rapidly-acquired low-resolution images into higher-resolution images. This work compares DL-CS with DL-SR for accelerated MRI on a test dataset of 50 patients with conventional image quality metrics and clinically-relevant quantitative T2 relaxation measurements. We demonstrate that DLCS approaches outperform DLSR approaches for accelerated MRI. |
| 1874 | Computer 54
|
Image Quality Transfer improves the potential clinical value of low-field MRI |
| Matteo Figini1,2, Hongxiang Lin1,2,3, Felice D'Arco4, Godwin Inalegwu Ogbole5, Maria Camilla Rossi Espagnet6, Olalekan Ibukun Oyinloye7, Joseph O Yaria8, Donald Amasike Nzeh7, Mojisola Omolola Atalabi9, Lisa Ronan1,2, David W Carmichael10,11, Judith Helen Cross4,11, Ikeoluwa A Lagunju12, Delmiro Fernandez-Reyes2,12, and Daniel C Alexander1,2 | ||
1Centre for Medical Image Computing, University College London, London, United Kingdom, 2Computer Science, University College London, London, United Kingdom, 3Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China, 4Great Ormond Street Hospital for Children, London, United Kingdom, 5Radiology, College of Medicine, University of Ibadan, Ibadan, Nigeria, 6Neuroradiology, Sapienza University, Rome, Italy, 7Radiology, University of Ilorin Teaching Hospital, Ilorin, Nigeria, 8Neurology, University College Hospital Ibadan, Ibadan, Nigeria, 9Radiology, University College Hospital Ibadan, Ibadan, Nigeria, 10School of Biomedical Engineering & Imaging Sciences, King's College London, London, United Kingdom, 11UCL Great Ormond Street Institute of Child Health, London, United Kingdom, 12Paediatrics, College of Medicine, University of Ibadan, Ibadan, Nigeria |
||
We applied Image Quality Transfer to enhance the contrast and resolution in the slice direction of low-field structural MRI, using a deep learning model trained on simulated images. Six radiologists blindly reviewed the enhanced images compared to low- and high-field images from 12 paediatric patients with epilepsy. Results demonstrated significant improvement of the visualisation of brain structures in sagittal and coronal orientations, and marginal improvement of the contrast between grey and white matter. If these promising results are confirmed in a larger study and in lesions, IQT could be an important tool to enhance the diagnostic power of low-field MRI. |
| 1875 | Computer 55
|
Uncertainty quantification for ground-truth free evaluation of deep learning reconstructions |
| Soumick Chatterjee1,2,3, Alessandro Sciarra1,4, Max Dünnwald3,4, Anitha Bhat Talagini Ashoka3, Mayura Gurjar Cheepinahalli Vasudeva3, Shudarsan Saravanan3, Venkatesh Thirugnana Sambandham3, Steffen Oeltze-Jafra4,5, Oliver Speck1,5,6,7, and Andreas Nürnberger2,3,5 | ||
1Department of Biomedical Magnetic Resonance, Otto von Guericke University Magdeburg, Magdeburg, Germany, 2Data and Knowledge Engineering Group, Otto von Guericke University Magdeburg, Magdeburg, Germany, 3Faculty of Computer Science, Otto von Guericke University Magdeburg, Magdeburg, Germany, 4MedDigit, Department of Neurology, Medical Faculty, University Hospital, Magdeburg, Germany, 5Center for Behavioral Brain Sciences, Magdeburg, Germany, 6German Centre for NeurodegenerativeDiseases, Magdeburg, Germany, 7Leibniz Institute for Neurobiology, Magdeburg, Germany |
||
Many deep learning-based techniques have been proposed in recent years to reconstruct undersampled MRI – showing their potential for shortening the acquisition time. Before using them in actual practice, they are usually evaluated by comparing their results against the available ground-truth – which is not available during real applications. This research shows the potential of using uncertainty estimation to evaluate the reconstructions without using any ground-truth images. The method has been evaluated for the task of super-resolution MRI, for acceleration factors ranging from two to four – in all three dimensions. |
| 1876 | Computer 56
|
On the importance of fetal brain numerical models for domain adaptation strategies in fetal brain MRI tissue segmentation |
| Priscille de Dumast1,2, Hamza Kebiri1,2, Meritxell Bach Cuadra1,2, and Hélène Lajous1,2 | ||
1CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 2Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland |
||
Manual fetal brain tissue segmentation is needed for training machine learning methods but is a tedious and error-prone task. The generation of synthetic magnetic resonance images can overcome the lack of clinical annotations by supplementing scarce clinical fetal datasets. However, we highlight that the choice of the numerical model from which additional data are derived is key to maximize the segmentation accuracy of clinical data via domain adaptation strategies. We demonstrate that the resort to high-resolution segmented images from real neurotypical and pathological cases enhances the morphological variability compared to an atlas, resulting in improved fetal brain tissue delineation overall. |
| 1877 | Computer 57
|
A deep learning-based quality control system for co-registration of prostate MR images |
| Mohammed R. S. Sunoqrot1, Kirsten M. Selnæs1,2, Bendik S. Abrahamsen1, Alexandros Patsanis1, Gabriel A. Nketiah1,2, Tone F. Bathen1,2, and Mattijs Elschot1,2 | ||
1Department of Circulation and Medical Imaging, Norwegian University of Science and Technology (NTNU), Trondheim, Norway, 2Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway |
||
Multiparametric MRI (mpMRI) is a valuable tool for the diagnosis of prostate cancer. Computer-aided detection and diagnosis (CAD) systems have the potential to improve robustness and efficiency compared to traditional radiological reading of mpMRI in prostate cancer. Co-registration of diffusion-weighted and T2-weighted images is a crucial step of CAD but is not always flawless. Automated detection of poorly co-registered cases would therefore be a useful supplement. This work shows that a fully automated quality control system for co-registration of prostate MR images based on deep learning has potential for this purpose. |
| 1878 | Computer 58
|
MR image enhancement using a multi-task neural network trained using only synthetic data |
| Kevin Blansit1, Zhehao Hu1, Greg Zaharchuk1, Enhao Gong1, and Keshav Datta1 | ||
1Subtle Medical, Menlo Park, CA, United States |
||
Faster MRI scans can be achieved by acquiring low resolution images or low SNR images and enhancing them to standard-of-care using deep learning techniques. However, to achieve clinical diagnostic quality images, this requires a large number of paired clinical datasets to train the model. Here we show that a multi-task deep convolutional neural network (DCNN) trained using only simulated motion artifact, low SNR, and low resolution images is capable of improving the quality of clinically acquired images from motion corrupted and accelerated sequences. |
| 1879 | Computer 59
|
The Utility of Virtual Biopsies for Dataset Augmentation as Applied to AI-Based Detection of Tumor Infiltration in Non-Enhancing Brain Lesion |
| Robert Wujek1, Melissa Prah2, Mona Al-Gizawiy2, and Kathleen Schmainda2 | ||
1Graduate School, Medical College of Wisconsin, Wauwatosa, WI, United States, 2Biophysics, Medical College of Wisconsin, Wauwatosa, WI, United States |
||
Delineation of invasive tumor from peritumoral edematous tissue remains a major obstacle to glioma treatment. To address this problem, a neural network was trained to distinguish between these regions using biopsies paired with colocalized MRI inputs. In addition to histologically confirmed biopsies, virtual biopsies sampled from non-contrast enhancing, FLAIR enhancing regions of non-invasive tumors (meningioma, metastasis) were used with an assumed classification of “non-tumor”. The current work is a preliminary assessment of this assumption and it's impact on model performance. |
| 1880 | Computer 60
|
QC of image registration using a DL network trained using only synthetic images |
| Yiheng Li1 | ||
1Subtle Medical, Santa Clara, CA, United States |
||
To perform robust quality control of medical imaging registration, we propose a method that can QC co-registration for multiple organs, without the restriction of the modalities. A rule-based image synthesis pipeline is used to generate random contrasts and shapes as training images. ResNet34 is trained to predict image alignment. Two MRI datasets with the spine or brain as subjects are used as external validation sets. The proposed model trained with synthetic images is validated on either one of the real MRI datasets and outperforms the same model trained on the other MRI dataset, which shows better generalizability. |
| 1881 | Computer 61
|
Paired CycleGAN-based Cross Vendor Harmonization |
| Joonhyeok Yoon1, Sooyeon Ji1, Eun-Jung Choi1, Hwihun Jeong1, and Jongho Lee1 | ||
1Seoul National University, Seoul, Korea, Republic of |
||
CycleGAN shows good performance with harmonization task. However, generative models have risk of structure modification. we proposed a cross-vendor harmonization model with paired CycleGAN based architecture for both high performance and structural consistency. We acquired 4 in-vivo dataset from Siemens and Philips scanner. For algorithm, we adapted CycleGAN for generation model, and we utilized L1 loss from pix2pix and patchGAN discriminator for structural consistency. Evaluations were performed both quantitatively and qualitatively. To quantitative evaluation, we assessed means of structural similarity index measure (SSIM). Proposed model shows better results compared to CycleGAN architecture. |
| 1882 | Computer 62
|
Inter-scanner harmonization of T1-weighted brain images using 3D CycleGAN |
| Vincent Roca1, Grégory Kuchcinski1,2, Morgan Gautherot1, Xavier Leclerc1,2, Jean-Pierre Pruvo1,2, and Renaud Lopes1,2 | ||
159000, Univ Lille, UMS 2014 – US 41 – PLBS – Plateformes Lilloises en Biologie & Santé, Lille, France, 259000, Univ Lille, Inserm, Lille Neuroscience & Cognition, Lille, France |
||
In MRI multicentric studies, inter-scanner harmonization is necessary to avoid taking into account variations due to technical differences in the analysis. In this study, we focused on CycleGAN models for 3D T1 weighted brain images harmonization. More precisely, we didn't follow the classical 2D CycleGAN architecure and developped a 3D cycleGAN model. We compared harmonization quality of these two kinds of models using 20 imaging features quantifying T1 signal, contrast between brain structures and segmentation quality. Results illustrate the potential of 3D CycleGAN for better synthesize images in inter-scanner MRI harmonization tasks. |
| 1950 | Computer 45
|
Motion tracking verification with a 3D printed compliant mechanisms and laser interferometry |
| Christoph Michael Schildknecht1 and Klaas Paul Prüssmann1 | ||
1Institute for Biomedical Engineering, ETH Zurich and University of Zurich, Zürich, Switzerland |
||
Precisely characterizing motion tracking modalities in the MRI bore is quite a challenge. Especially when not only static positions ought to be evaluated, but also dynamic characterization is desired. In this work, we show how a compliant mechanism in the form of a folded-beam suspension can be utilized for accurate dynamic positioning. A piezoelectric actuator drives the system, and position feedback is acquired with a laser interferometer. Testing the system with short-wave motion tracking revealed a sub-micrometre agreement. |
| 1951 | Computer 46
|
Evaluation of markerless prospective motion correction for neuroanatomical MRI |
| Zakaria Zariry1,2, Robert Frost3,4, Franck Lamberton2,5, Thomas Troalen6, Nathalie Richard1,2, Andre van der Kouwe3,4, and Bassem Hiba1,2 | ||
1Institute of Cognitive Neuroscience Marc Jeannerod, CNRS / UMR 5229, BRON, France, Metropolitan, 2Université Claude Bernard - Lyon 1, Villeurbanne, France, Metropolitan, 3Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 4Department of Radiology, Harvard Medical School, Boston, MA, United States, 5Cermep, CNRS / UAR 3453, Lyon, France, Metropolitan, 6Siemens Healthcare SAS, Saint-Denis, France, Metropolitan |
||
Head-motion is a main cause of artifacts in MRI. The performance of a markerless optical system, which records the subject's face, estimates head-motion, and allows real-time repositioning of the FOV, is evaluated for neuroanatomical MRI. Sets of T1W/T2W-images were collected from 2 subjects instructed to perform different head-motion protocols during the acquisitions. The System performance is evaluated by comparing images with/without motion-correction. The optical system ensures good corrections of head-motions. However, the correction quality depends on the amplitude of the movement, its location in k-space and its nature. Residual effects of large amplitudes/amounts movements may persist on corrected images. |
| 1952 | Computer 47
|
Think outside the box: Exploiting the imaging workflow for Deep Learning based motion estimation and correction |
| Julian Hossbach1,2, Daniel Nicolas Splitthoff3, Bryan Clifford4, Daniel Polak3,5, Wei-Ching Lo4, Stephan Cauley5, and Andreas Maier1 | ||
1Pattern Recognition Lab Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany, 2Erlangen Graduate School in Advanced Optical Technologies, Erlangen, Germany, 3Siemens Healthcare, Erlangen, Germany, 4Siemens Medical Solutions, Malden, MA, United States, 5Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States |
||
Estimating intrascan subject motion enables the reduction of motion artifacts but often requires further calibration data e.g. from an additional motion-free reference. In this work, we explore how the reuse of supplementary scans in the imaging workflow can be used as motion calibration data. More specifically, the preceding parallel imaging calibration scan is reutilized to support a Deep Learning (DL) approach for estimating motion. Results are presented which indicate that DL, in contrast to a conventional optimization approach, can extract the motion and improve the image quality despite contrast differences between the calibration and imaging scans. |
| 1953 | Computer 48
|
Joint Neural Network for Fast Retrospective Rigid Motion Correction of Accelerated Segmented Multislice MRI |
| Nalini M Singh1,2, Malte Hoffmann3,4, Daniel C Moyer1, Ikbeom Jang3,4, Lina Chen5, Marcio Aloisio Bezerra Cavalcanti Rockenbach5, Arnaud Guidon6, Iman Aganj3,4, Jayashree Kalpathy-Cramer3,4,5, Elfar Adalsteinsson2,7,8, Bruce Fischl2,3,4, Adrian V Dalca1,3,4, Polina Golland*1,7,8, and Robert Frost*3,4 | ||
1Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, United States, 2Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 3Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, United States, 4Department of Radiology, Harvard Medical School, Boston, MA, United States, 5MGH & BWH Center for Clinical Data Science, Mass General Brigham, Boston, MA, United States, 6GE Healthcare, Boston, MA, United States, 7Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 8Institute for Medical Engineering and Science, Massachusetts Institute of Technology, Cambridge, MA, United States |
||
We demonstrate a deep learning approach for fast retrospective intraslice rigid motion correction in segmented multislice MRI. The proposed neural network architecture combines convolutions on frequency and image space representations of the acquired data to produce high quality reconstructions. Unlike many prior techniques, our method does not require auxiliary information on the subject head motion during the scan. The resulting reconstruction procedure is more accurate and is an order of magnitude faster than GRAPPA. Our work offers the first step toward fast motion correction in any setting with substantial, unpredictable, challenging to track motion. |
| 1954 | Computer 49
|
Accelerated 3D EPI navigator for prospective motion correction |
| Yulin Chang1, Daniel Nicolas Splitthoff2, Wei-ching Lo1, M. Dylan Tisdall3, and Andre van der Kouwe4 | ||
1Siemens Medical Solutions USA Inc., Malvern, PA, United States, 2Siemens Healthcare GmbH, Erlangen, Germany, 3Radiology, University of Pennsylvania, Philadelphia, PA, United States, 4Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Harvard Medical School, Massachusetts General Hospital, Boston, MA, United States |
||
We show that for navigator-based prospective motion correction MRI, acceleration of 3D EPI acquisition increases sequence flexibility and improves the navigator image quality without sacrificing the quality of motion correction. |
| 1955 | Computer 50
|
Pilot-Tone Motion Estimation for Brain Imaging at Ultra-High Field Using FatNav Calibration |
| Tom Wilkinson1,2, Yannick Brackenier1,2, Felipe Godinez3, Raphael Tomi-Tricot1,2,4, Jan Sedlacik1,2, Philippa Bridgen1,2, Sharon Giles1,2, Joseph V Hajnal1,2, and Shaihan J Malik1,2 | ||
1Centre for the Developing Brain, School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2London Collaborative Ultra high field System (LoCUS), London, United Kingdom, 3Department of Radiology, University of California Davis, Sacramento, CA, United States, 4MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom |
||
3D FatNavs were used to provide a rapid, robust pre-scan to calibrate motion estimation from pilot-tone. The amount of training data required to make reliable forward predictions was investigated. This method robustly predicts motion within a dataset and can be applied to other datasets with lower accuracy. Improving the accuracy and speed of this method is ongoing work. This independent method of motion estimation shows promise for rapid calibration as part of routine examinations and paves the way to offer motion correction at ultra-high field where motion-correction is particularly relevant. |
| 1956 | Computer 51
|
Prospective motion correction in multi-inversion EPI using volumetric navigators for robust T1 map estimation |
| Jonathan R. Polimeni1,2,3, M. Dylan Tisdall4, Daniel J. Park1, Paul Wighton1, S. Robert Frost1,2, Christine L. Tardif5,6, and Andre J. W. van der Kouwe1,2 | ||
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 2Department of Radiology, Harvard Medical School, Boston, MA, United States, 3Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 4Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States, 5Department of Biomedical Engineering, Department of Neurology & Neurosurgery, McGill University, Montreal, QC, Canada, 6McConnell Brain Imaging Centre, Montreal Neurological Institute, Montreal, QC, Canada |
||
In multi-inversion EPI (MI-EPI), each slice samples a distinct inversion time during each inversion recovery, providing an efficient method for estimating T1. MI-EPI is vulnerable to through-plane motion, which results in slices sampling a subset of the desired inversion times and wrong TIs will be attributed to the slices. This cannot be corrected retrospectively. We introduce prospective motion correction in MI-EPI using volumetric navigators (vNavs). vNavs are acquired at the beginning of the inversion recovery thus the effects of their excitation pulses must be modeled for T1 estimation. This provides improved T1 estimation accuracy in the presence of subject motion. |
| 1957 | Computer 52
|
Retrospective motion correction with structural priors for clinical MRI protocols |
| Gabrio Rizzuti1, Niek Huttinga2, Alessandro Sbrizzi2, and Tristan van Leeuwen3 | ||
1Utrecht University, Utrecht, Netherlands, 2Universitair Medisch Centrum Utrecht, Utrecht, Netherlands, 3Centrum Wiskunde & Informatica, Amsterdam, Netherlands |
||
We present a retrospective rigid motion correction and reconstruction scheme for brain MRI with the aid of structural priors. The proposed framework is designed to be applied to a clinical protocol including multiple scans for multiple contrasts, of which some scans are corrupted due to motion during the acquisitions, but at least one is uncorrupted and exploited as an additional prior. We envision a practical workflow that can easily fit into the existing clinical practice without the need for changing the MRI acquisition protocols. |
| 1958 | Computer 53
|
Gaussian Process Modelling and Compensation of Motion in DCE-MRI |
| Aziz Kocanaogullari1, Cemre Ariyurek1, Onur Afacan1, and Sila Kurugol1 | ||
1Radiology, Boston Children's Hospital, Boston, MA, United States |
||
Radial DCE-MRI is robust to motion. However, bulk motion or heavy breathing causes 1) irrecoverably deteriorated k-space lines acquired during motion events reducing image quality, 2) misaligned volumes in a dynamic sequence. In this work we propose to solve the first problem by fitting a Gaussian process to the k-space center of each spoke over time and using it to determine outlier spokes corrupted by motion. We solve the second problem by clustering the dynamic data to respective motionless phases before and after each motion event and registering volumes between phases for computationally efficient correction of motion with fewer registrations. |
| 1959 | Computer 54
|
Self-Assisted Priors with Cascaded Refinement Network for Reduction of Rigid Motion Artifacts in Brain MRI |
| Mohammed A. Al-masni1, Seul Lee1, Jaeuk Yi1, Sewook Kim1, Sung-Min Gho2, Young Hun Choi3, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, College of Engineering, Yonsei University, Seoul, Korea, Republic of, 2GE Healthcare, Seoul, Korea, Republic of, 3Department of Radiology, Seoul National University Hospital, Seoul, Korea, Republic of |
||
MRI is sensitive to motion caused by patient movement. It may cause severe degradation of image quality. We develop an efficient retrospective deep learning method called stacked U-Nets with self-assisted priors to reduce the rigid motion artifacts in MRI. The proposed work exploits the usage of additional knowledge priors from the corrupted images themselves without the need for additional contrast data. We further design a refinement stacked U-Nets that facilitates preserving of the image spatial details and hence improves the pixel-to-pixel dependency. The experimental results prove the feasibility of self-assisted priors since it does not require any further data scans. |
| 1960 | Computer 55
|
Registration of abdominal DCE-MRI: optimal choice of reference and moving image pairs |
| Adam George Tattersall1,2, Keith A Goatman2, Scott Semple1, and Lucy E Kershaw1 | ||
1University of Edinburgh, Edinburgh, United Kingdom, 2Canon Medical Research Europe, Edinburgh, United Kingdom |
||
Before registration can take place, an approach to pairing images must be chosen. We used synthetic data to provide a ground truth to evaluate four approaches to pairing images. Our synthetic data was created using tracer kinetics modelling of abdominal DCE-MRI to create motionless data. Then local distortions were applied to the motionless data. We found that using one image as a reference image produced the best result to eliminate the propagation of errors. The reference image must have a similar intensity distribution to the other images in the sequence. |
| 1961 | Computer 56
|
Automatic 3D to 2D reformatting in 4D flow MRI using continuous reinforced learning |
| Javier Bisbal1,2,3, Julio Sotelo1,3,4, Cristobal Arrieta1,3, Pablo Irarrázabal1,2,3, Marcelo Andia1,3,5, Cristian Tejos1,2,3, and Sergio Uribe1,3,5 | ||
1Biomedical Imaging Center UC, Pontificia Universidad Catolica de Chile, Santiago, Chile, 2Electrical Engineering Department, School of Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, 3ANID – Millennium Science Initiative Program – Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, 4School of Biomedical Engineering, Universidad de Valparaíso, Valparaíso, Chile, 5Department of Radiology, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile |
||
One major limitation on 4D flow MRI is the time-consuming and user-dependent post-processing. We developed an automated reinforced deep learning framework for plane planning in 4D flow data. This method sequentially updates plane parameters towards a target plane based on a continuous policy. A total of 83 4D flow MRI scans were considered, 41 for training, 14 for validation and 28 for test. Our method achieves good results in terms of angulation and distance error (9.21 ± 3.85 degrees and 3.72 ± 2.19 mm). |
| 1962 | Computer 57
|
Single-image superresolution of hyperpolarized 13C spectroscopic images |
| Kofi Deh1, Nathaniel Kim1, Guannan Zhang1, Miloushev Vesselin1, and Kayvan Keshari1 | ||
1Memorial Sloan Kettering Cancer Center, New York, NY, United States |
||
Hyperpolarized 13C spectroscopic images are acquired at a low spatial resolution, making it necessary to apply superresolution techniques to the metabolite maps prior to fusion with the proton anatomic image for visualization of metabolite biodistribution. In this work, we demonstrate great improvement in image quality for preclinical and clinical images when the metabolite map is upsampled by high spatial frequency transfer from magnetic resonance images of thermally polarized 13C phantoms to the invivo metabolic image. |
| 1963 | Computer 58
|
Learning motion correction from YouTube for real-time MRI reconstruction with AUTOMAP |
| David E J Waddington1, Christopher Chiu1, Nicholas Hindley1,2, Neha Koonjoo2, Tess Reynolds1, Paul Liu1, Bo Zhu2, Chiara Paganelli3, Matthew S Rosen2,4,5, and Paul J Keall1 | ||
1ACRF Image X Institute, Faculty of Medicine and Health, The University of Sydney, Sydney, Australia, 2A. A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, 3Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano, Milan, Italy, 4Department of Physics, Harvard University, Cambridge, MA, United States, 5Harvard Medical School, Boston, MA, United States |
||
Today’s MRI does not have the spatio-temporal resolution to image the anatomy of a patient in real-time. Therefore, novel solutions are required in MRI-guided radiotherapy to enable real-time adaptation of the treatment beam to optimally target the cancer and spare surrounding healthy tissue. Neural networks could solve this problem, however, there is a dearth of sufficiently large training data required to accurately model patient motion. Here, we use the YouTube-8M database to train the AUTOMAP network. We use a virtual dynamic lung tumour phantom to show that the generalized motion properties learned from YouTube lead to improved target tracking accuracy. |
| 1964 | Computer 59
|
MR-Class: MR Image Classification using one-vs-all Deep Convolutional Neural Network |
| Patrick Salome1,2,3,4, Francesco Sforazzini1,2,3,4, Andreas Kudak3,5,6, Matthias Dostal3,5,6, Nina Bougatf3,5,6, Jürgen Debus3,4,5,7, Amir Abdollahi1,3,4,5, and Maximilian Knoll1,3,4,5 | ||
1CCU Translational Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany, 2Medical Faculty, Heidelberg University Hospital, Heidelberg, Germany, 3Heidelberg Ion-Beam Therapy Center (HIT), Heidelberg, Germany, 4German Cancer Consortium (DKTK) Core Center, Heidelberg, Germany, 5Radiation Oncology, Heidelberg University Hospital, Heidelberg, Germany, 6CCU Radiation Therapy, German Cancer Research Center (DKFZ), Heidelberg, Germany, 7National Center for Tumor Diseases (NCT), Heidelberg, Germany |
||
MR-Class is a deep learning-based MR image classification tool that facilitates and speeds up the initialization of big data MR-based research studies by providing fast, robust, and quality-assured MR image classifications. It was observed in this study that corrupt and misleading DICOM metadata could lead to a misclassification of about 10%. Therefore, in a field where independent datasets are frequently needed for study validations, MR-Class can eliminate the cumbrousness of data cohorts curation and sorting. This can greatly impact researchers interested in big data multiparametric MRI studies and thus contribute to the faster deployment of clinical artificial intelligence applications. |
| 1965 | Computer 60
|
Medical Image Registration Using Deep Learning Techniques Applied to Pediatric Magnetic Resonance Imaging (MRI) Brain Scans |
| Andjela Dimitrijevic1,2, Vincent Noblet3, and Benjamin De Leener1,2,4 | ||
1NeuroPoly Lab, Institute of Biomedical Engineering, Polytechnique Montréal, Montréal, QC, Canada, 2Research Center, Ste-Justine Hospital University Centre, Montréal, QC, Canada, 3ICube-UMR 7357, Université de Strasbourg, CNRS, Strasbourg, France, 4Computer Engineering and Software Engineering, Polytechnique Montréal, Montréal, QC, Canada |
||
Deep learning techniques have a potential in allowing fast deformable registration tasks. Studies around registration often focus on adult populations, even if there is a need for pediatric research where less data and studies are being produced. In this study, we compared three methods for intra-subject registration on publicly available Calgary Preschool dataset. Using the DeepReg framework, pre-registering with a rigid and affine transformation (proposed RigidAffineReg method) showed the least negative JD values and the highest Dice score (0.924±0.045). By achieving faster alignments, this tool for pediatric MRI scans could help proliferate larger scale population research in brain developmental studies. |
| 1966 | Computer 61
|
Deep Learning with Perceptual Loss Enables Super-Resolution in 7T Diffusion Images of the Human Brain |
| David Lohr1, Theresa Reiter1,2, and Laura Schreiber1 | ||
1Chair of Cellular and Molecular Imaging, Comprehensive Heart Failure Center (CHFC), Würzburg, Germany, 2Department of Internal Medicine I, University Hospital Würzburg, Würzburg, Germany |
||
Diffusion weighted imaging has become a key imaging modality for the assessment of brain connectivity as well as the structural integrity, but the method is limited by low SNR and long scan times. In this study we demonstrate that AI models trained on a moderate number of publicly available 7T datasets (n=12) are able to enhance image resolutions in diffusion MRI up to 25 times. The applied NoGAN model performs well for smaller resolution enhancements (4- and 9-fold), but generates "hyper" realistic images for higher enhancements (16- and 25-fold). Models trained using perceptual loss seem to avoid this limitation. |
| 1967 | Computer 62
|
Longitudinal Multiple Sclerosis Lesion Segmentation Using Pre-activation U-Net |
| Pooya Ashtari1,2, Berardino Barile1,2, Dominique Sappey-Marinier2, and Sabine Van Huffel1 | ||
1Department of Electrical Engineering (ESAT), KU Leuven, Leuven, Belgium, 2CREATIS (CNRS UMR5220 & INSERM U1294), Université Claude-Bernard Lyon 1, Lyon, France |
||
Automated segmentation of new multiple sclerosis (MS) lesions in MRI data is crucial for monitoring and quantifying MS progression. Manual delineation of such lesions is laborious and time-consuming since experts need to deal with 3D images and numerous small lesions. We propose a 3D encoder-decoder architecture with pre-activation blocks to segment new MS lesions in longitudinal FLAIR images. We also applied intensive data augmentation and deep supervision to mitigate the limited data and the class imbalance problem. The proposed model, called Pre-U-Net, achieved a Dice score of 0.62 and a sensitivity of 0.58 on the public challenge MSSEG-2 dataset. |
| 1968 | Computer 63
|
A Mask Guided Attention Generative Adversarial Network for Contrast-enhanced T1-weight MR Synthesis |
| Yajing Zhang1, Xiangyu Xiong1, and Chuanqi Sun1 | ||
1MR Clinical Science, Philips Healthcare, Suzhou, China |
||
Image synthesis methods based on deep learning has recently achieved success in reducing the dosage of gadolinium-based contrast agents (GBCAs). However, these methods cannot focus on the region of interest to synthesize realistic images. To address this issue, a mask guided attention generative adversarial network (MGA-GAN) was proposed to synthesize contrast enhanced T1-weight images from the multi-channel inputs. Qualitive and quantitative results indicate that the proposed MGA-GAN can improve the synthesized images with higher quality for details of brainstem glioma, compared with state-of-the-art methods. |
| 1969 | Computer 64
|
Retrospective correction of B1 field inhomogeneities in T2w 7T prostate patient data |
| Seb Harrevelt1, Daan Reesink2, Astrid Lier, van3, Richard Meijer3, Josien Pluim4, and Alexander Raaijmakers1 | ||
1TU Eindhoven, Utrecht, Netherlands, 2Meander Medisch Centrum, Utrecht, Netherlands, 3UMC Utrecht, Utrecht, Netherlands, 4TU Eindhoven, Eindhoven, Netherlands |
||
Prostate imaging at ultra-high fields is heavily affected by B1 field induced inhomogeneities. This not only results in unattractive images but it also might affect clinical diagnosis . To remedy this we developed a deep learning model that retrospectively corrects for the bias field. We applied this model to a clinical data set and demonstrated its performance in a qualitative manner. The results indicate that the model is able to drastically reduce the inhomogeneities in a variety of cases while the tissue contrast is generally maintained and the underlying anatomy has been successfully recovered. |
| 1970 | Computer 65
|
EPI Nyquist Ghost Artifact Correction for Brain Diffusion Weighted Imaging (DWI) using Deep Learning |
| Fatima Sattar1, Sadia Ahsan1, Fariha Aamir1, Ibtisam Aslam1,2, Iram Shahzadi3,4, and Hammad Omer1 | ||
1Medical Image Processing Research Group (MIPRG), Department of Electrical and Computer Engineering, COMSATS University, Islamabad, Pakistan, 2Service of Radiology, Geneva University Hospitals and Faculty of Medicine, Hospital University of Geneva and University of Geneva, Geneva, Switzerland, 3OncoRay – National Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Helmholtz-Zentrum Dresden –Rossendorf, Dresden, Germany, Dresden, Germany, 4German Cancer Research Center (DKFZ), Heidelberg, Germany, Dresden, Germany |
||
Echo-planar imaging suffers from Nyquist ghost (i.e., N/2 ghost) artifacts because of poor system gradients and delays. Many conventional methods have been used in literature to remove N/2 artifacts in Diffusion Weighted Imaging (DWI) but often produce erroneous results. This paper presents a deep learning approach to eliminate the phase error of k-space for removing the Nyquist ghost artifacts in DWI. Experimental results show successful removal of the ghost artifacts with improved SNR and reconstruction quality with the proposed method. |
| 1971 | Computer 66
|
Residual U-net for denoising 3D low field MRI |
| Tobias Senft1, Reina Ayde1, Marco Fiorito1, Najat Salameh1, and Mathieu Sarracanie1 | ||
1Department of Biomedical Engineering, University of Basel, Allschwil, Switzerland |
||
Low magnetic field (LF) MRI is currently gaining momentum as a complementary, more flexible and cost-effective approach to MRI diagnosis. However, the impaired signal-to-noise ratio challenges its relevance for clinical applications. Recently, denoising of low SNR images using deep learning techniques has shown promising results for MRI applications. In this study, we assess the denoising performance of residual U-net architecture on different SNR levels of LF MRI data (0.1 T). The model performance has been evaluated on both simulated and acquired LF MRI datasets. |
| 1972 | Computer 67
|
Artificially-generated Consolidations and Balanced Augmentation increase Performance of U-Net for Lung Parenchyma Segmentation on MR Images |
| Cristian Crisosto1,2, Andreas Voskrebenzev1,2, Marcel Gutberlet1,2, Filip Klimeš1,2, Till Kaireit1,2, Gesa Pöhler1,2, Tawfik Alsady1,2, Lea Behrendt1,2, Robin Müller1,2, Maximilian Zubke1,2, Frank Wacker1,2, and Jens Vogel-Claussen1,2 | ||
1Institute of Diagnostic and Interventional Radiology, Medical School Hannover, Hannover, Germany, 2Biomedical Research in Endstage and Obstructive Lung Disease Hannover (BREATH), Member of the German Centre for Lung Research (DZL), Hannover, Germany, Hannover, Germany |
||
Accurate fully automated lung segmentation is needed to facilitate Fourier-Decomposition employment-based techniques in clinical routine among different centers. However, the lung parenchyma segmentation remains challenging for convolutional neural networks (CNN) when consolidations are present. To improve training balanced augmentation (BA) and artificially-generated consolidations (AGC) were introduced. The proposed CNN was compared to conventional CNNs without BA and AGC using Sørensen-Dice coefficient (SDC) and Hausdorff coefficient (HD). The SDC / HD of the proposed model is significantly higher (p of 0.0001 and p of 0.0146 / p of 0.0009 and p of 0.0152) when compared to CNNs without BA and AGC. |
| 1973 | Computer 68
|
Subject classification based on functional connectivity and white matter microstructure in a rat model of Alzheimer’s using machine learning |
| Yujian Diao1,2,3, Catarina Tristão Pereira2,4, Ting Yin2, and Ileana Ozana Jelescu2,5 | ||
1Laboratory of Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 2CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 3Animal Imaging and Technology, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 4Faculdade de Ciências da Universidade de Lisboa, Lisbon, Portugal, 5Department of Radiology, Lausanne University Hospital, Lausanne, Switzerland |
||
Impaired brain glucose consumption is a possible trigger of Alzheimer’s disease (AD). Previous work revealed affected brain structure and function by insulin resistance in terms of functional connectivity and white matter microstructure in a rat model of AD. Here, functional and structural metrics were further used to classify Alzheimer’s from control rats using logistic regression. Our study highlights the MRI-derived biomarkers that best discriminate Alzheimer’s vs control rats early in the course of the disease, with potential translation to human AD. |
| 1974 | Computer 69
|
Mapping human brain development at new spatial resolutions using deep learning and high-resolution quantitative MRI |
| Georgia Doumou1,2, Hongxiang Lin2,3, Sara Lorio1, Lenka Vaculčiaková4, Kerrin J. Pine4, Nikolaus Weiskopf4,5, Jonathan O'Muircheartaigh6,7,8, Daniel C. Alexander2, and David W. Carmichael 1 | ||
1Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, 3Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China, 4Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 5Felix Bloch Institute for Solid State Physics, Faculty of Physics and Earth Sciences, Leipzig University, Leipzig, Germany, 6Department of Forensic & Neurodevelopmental Sciences, King's College London, London, United Kingdom, 7Centre for the Developing Brain, Department of Perinatal Imaging and Health, King's College London, London, United Kingdom, 8MRC Centre for Neurodevelopmental Disorders, King's College London, London, United Kingdom |
||
High-resolution quantitative MRI using ultra-high field scanners (7T) could advance a range of research and clinical applications if limitations, both practical (e.g. long acquisitions) and technical (e.g. B1 non-uniformity), can be avoided. This could be achieved by using 7T information to enhance conventional field strength images. To test this approach, paired 3T-7T R1 maps were used to train a U-Net variant to enhance 3T R1 maps. Leave-one-out cross-validation, quantitative evaluation, as well as external validation on an external clinical dataset, demonstrated promising enhancement with visual and quantitative metrics more similar to 7T R1 maps than the 3T equivalents. |
| 1975 | Computer 70
|
3D brain MRI synthesis utilizing 2D SPADE-GAN and 3D CNN architecture |
| Aymen Ayaz1, Ronald de Jong1, Samaneh Abbasi-Sureshjani1, Sina Amirrajab1, Cristian Lorenz2, Juergen Weese2, and Marcel Breeuwer1,3 | ||
1Biomedical Engineering Department, Eindhoven University of Technology, Eindhoven, Netherlands, 2Philips Research Laboratories, Hamburg, Germany, 3MR R&D – Clinical Science, Philips Healthcare, Best, Netherlands |
||
We propose a method to synthesize 3D consistent brain MRI utilizing multiple 2D conditional SPatially-Adaptive (DE)normalization (SPADE) GANs to preserve the spatial information of patient-specific brain anatomy and a 3D CNN based network to improve the image consistency in all directions in a cost-efficient manner. Three individual 2D SPADE-GAN networks are trained across the axial, sagittal and coronal slice directions and their outputs are thereafter used further to train a CNN based 3D image restoration network to combine them into one 3D volume, using real MRI as a reference. The resulting predicted-synthesized 3D brain MRI is evaluated quantitatively and qualitatively. |
| 1976 | Computer 71
|
Quantitative MRI to characterize diffusion-controlled release of gadolinium from a calcium sulphate matrix |
| Greg Hong1,2,3, Tina Khazaee1,2,3, Santiago F. Cobos1,2,3, Spencer D. Christiansen1,2, Junmin Liu2, Maria Drangova1,2,3, and David W. Holdsworth1,2,3 | ||
1Medical Biophysics, Western University, London, ON, Canada, 2Robarts Research Institute, London, ON, Canada, 3Bone & Joint Institute, London, ON, Canada |
||
Calcium sulfate (CS) is commonly used to deliver antibiotics to treat periprosthetic joint infection, which is a leading cause of early revision. There is an unmet need for non-invasive measurement of antibiotic release from CS, which would improve understanding of antibiotic delivery in-vivo. Through the use of a gadolinium-based contrast agent as a surrogate, this study shows that quantitative MRI (R2* and QSM) can be used to track diffusion-controlled release. We demonstrate this in a phantom study consisting of a cylindrical CS core surrounded by agar, where gadolinium diffuses out of the core and through the agar sample. |
| 1977 | Computer 72
|
Use of MRI- and microscopy-compatible bioreactor for validation of qMRI diffusion measurements via fluorescence microscopy |
| Megan F LaMonica1, Megan E Poorman2, David A Hormuth, II1, Thomas E Yankeelov1, and Kathryn E Keenan2 | ||
1Biomedical Engineering, UT Austin, Austin, TX, United States, 2NIST, Boulder, CO, United States |
||
Quantitative magnetic resonance imaging (qMRI) could complement qualitative MRI analysis by providing quantitative diagnostic information standardized across patients. However, it is difficult to determine how microscopic tissue properties affect MR signal. A microscopy- and MRI-compatible bioreactor was filled with different concentrations of fluorescent beads in gel to explore this relationship. Higher bead count was associated with lower ADC. A 59% ADC decrease was observed at 100 beads/mm3 when varying gradient pulse separation from 16 to 20 ms, illustrating sensitivity to different length scales per the Einstein-Smoluchowski relationship. This work is a step towards validating qMRI with fluorescence microscopy. |
| 1978 | Computer 73
|
Advanced diffusion-weighted MRI and proteomic analysis as potential predictors of renal tumor histopathology |
| Octavia Bane1,2, Jorge Daza3, Berengere Salome4, John Sfakianos3, Andrew Charap4, Kirolos Meilika3, Kennedy Okhawere3, Amir Horowitz4, Bachir Taouli1,2, Ketan Badani3, and Sara Lewis1,2 | ||
1Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 3Department of Urology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 4Division of Oncological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, United States |
||
In this prospective, single-center study, we evaluated the use of intravoxel incoherent motion model of diffusion (IVIM-DWI) and high-throughput urine proteomics associated with inflammation for diagnosis and characterization of small renal masses. IVIM-DWI parameters had good to excellent diagnostic performance in identifying clear-cell renal cell carcinomas (ccRCC), as well as in distinguishing between stages of ccRCC. In a subset of patients with IVIM-DWI and proteomics, we observed significant correlations between IVIM and proteomic analytes which had differential expression between ccRCC and non-ccRCC, and are involved in promoting the recruitment, expansion and activation of innate and adaptive immune cells. |
| 1979 | Computer 74
|
Microstructure Imaging of Head and Neck Cancers using an Oscillating Diffusion Gradient Sequence and a Random Walk with Barriers Model |
| Siamak Nejad-Davarani1, Michelle Mierzwa1, Thorsten Feiweier2, Keith Casper3, and Yue Cao1,4 | ||
1Department of Radiation Oncology, University of Michigan, Ann Arbor, MI, United States, 2Siemens Healthcare GmbH, Erlangen, Germany, 3Department of Otolaryngology, University of Michigan, Ann Arbor, MI, United States, 4Department of Radiology, University of Michigan, Ann Arbor, MI, United States |
||
In vivo microstructure imaging based on diffusion weighted imaging can provide valuable information on tumor characteristics and response to cancer therapy. An oscillating diffusion gradient spin-echo sequence was used to acquire DW images with short diffusion times, along with those acquired by a pulsed diffusion gradient spin-echo sequence. Using a random walk with barriers model, microstructure/diffusion parameters in primary and nodal head and neck tumors were estimated prior to, and after two weeks of cancer treatment. Results show feasibility of this method to detect significant changes in these parameters, across the two imaging time points. |
| 1980 | Computer 75
|
Erwin - a quantitative MRI toolbox |
| Julien Lamy1, Mary Mondino1, and Paulo Loureiro de Sousa1 | ||
1ICube, Université de Strasbourg-CNRS, Strasbourg, France |
||
Erwin is an open-source Python toolbox dedicated to the computation of quantitative maps from MRI data. It provides a unified interface, through either its Python API or its command-line tool, to well-known qMRI methods (e.g. B1 mapping using Actual Flip Angle (AFI) or T1 mapping using Variable Flip Angle (VFA)) and external toolboxes (e.g. MRtrix or MEDI). Erwin also provides tools to create complex pipelines where the dependencies between steps are tracked: a change in a source image or a method parameter will only trigger the recomputation of a minimal set of subsequent steps. |
| 1981 | Computer 76
|
dtiRIM: A recurrent inference machine for diffusion tensor estimation |
| Emanoel R. Sabidussi1, Stefan Klein1, Ben Jeurissen2, and Dirk H. J. Poot1 | ||
1Radiology, Erasmus Medical Center, Rotterdam, Netherlands, 2Department of Physics, Imec-Vision Lab, University of Antwerp, Antwerp, Belgium |
||
In Diffusion MRI, the large variability of acquisition schemes limits broader use of Deep Learning for parameter estimation, with data-specific models needed for high-quality predictions. To reduce dependency on training data, we present dtiRIM, a recurrent neural network that learns a regularized solution to a model-based inverse problem. Using the diffusion model allows independent parameters (e.g. gradient directions) to also influence the estimation. We show that a single dtiRIM model predicts diffusion parameters for multiple datasets with lower error than the current state-of-the-art. Our study suggests that dtiRIM has the potential to be the first general learning method for DTI. |
| 1982 | Computer 77
|
Quantitative Susceptibility Mapping (QSM) Using High-resolution Ultra-Short Echo Time (UTE) MRI with Rosette k-space Pattern |
| Aparna Karnik1, Xin Shen2, Humberto Monsivais3, Antonia Sunjar2, Ali Caglar Özen4, Serhat Ilbey4, Mark Chiew5, Mukerrem Cakmak6, Joseph Rispoli1,2, and Uzay Emir3 | ||
1Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, United States, 2Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States, 3School of Health Sciences, Purdue University, West Lafayette, IN, United States, 4Department of Radiology, Medical Physics, Medical Center – University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, 5Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom, 6School of Materials Engineering, Purdue University, West Lafayette, IN, United States |
||
We developed a method of acquiring high-resolution MRI for the generation of Quantitative Susceptibility Mapping (QSM) using Ultra-short echo time (UTE) MRI with a novel 3D rosette k-space trajectory. This method was used to generate high-resolution (0.94 mm3 isotropic) magnetic susceptibility maps in an Iron-Chloride phantom and the human brain. Generated maps were then compared with the susceptibility maps reconstructed from low-resolution data acquired using multi-echo UTE acquisition. Susceptibility values were comparable with current literature demonstrating the promise of this novel method in the generation of QSM. |
| 1983 | Computer 78
|
Development of a Quantitative Susceptibility Mapping (QSM) phantom for validation of acquisition strategies and post-processing tools at 3 T |
| Padriac Hooper1,2, Monique Tourell1,2, Kieran O'Brien2,3, Jin Jin2,3, Simon Daniel Robinson1,4,5,6, and Markus Barth1,2,7 | ||
1Centre for Advanced Imaging, The University of Queensland, Brisbane, Australia, 2ARC Training Centre for Innovation in Biomedical Imaging Technology, Brisbane, Australia, 3Siemens Healthcare Pty Ltd, Brisbane, Australia, 4High Field MR Center, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria, 5Karl Landsteiner Institute for Clinical Molecular MR in Musculoskeletal Imaging, Vienna, Austria, 6Department of Neurology, Medical University of Graz, Graz, Austria, 7School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane, Australia |
||
The phantom was developed to cover the full range of physiological and pathological magnetic susceptibility seen in iron- and calcium-containing tissue, using 4 materials within the one phantom. The phantom facilitates quality assurance for acquisition strategies and post-processing tools. |
| 1984 | Computer 79
|
PedQ-NET : Unsupervised Model-Based Joint-Loss Deep learning for Pediatric QSM |
| Siyun Jung1, Soohyun Jeon1, Yoonho Nam2, Hyun Gi Kim3, and Dong-Hyun Kim1 | ||
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of, 2Division of Biomedical Engineering, Hankuk University of Foreign Studies, Yongin, Korea, Republic of, 3Department of Radiology, Catholic University of Korea, Eunpyeong St. Mary's Hospital, Seoul, Korea, Republic of |
||
In conventional QSM methods, it is difficult to analyze gray matter subcortical in pediatrics, because of small contrast difference between gray matter and white matter. In addition, there are no deep learning approaches that reflect the characteristics of the pediatric brain. In this study, we propose an unsupervised model-based joint-loss deep learning network for pediatric QSM, PedQ-NET. The proposed method achieved better edge preservation on the deep gray matter subcortical areas in pediatric QSM. According to the evaluation of in-vivo pediatric data, QSM generated by PedQ-NET shows enhanced edges in the subcortical areas and clear details with reduced artifacts. |
| 1985 | Computer 80
|
Realistic in silico abdominal QSM phantom |
| Javier Silva1,2,3, Carlos Milovic4, Mathias Lambert1,2,3, Cristian Montalba2,3, Cristobal Arrieta1,2,3, Sergio Uribe2,3,5, and Cristian Tejos1,2,3 | ||
1Department of Electrical Engineering, Pontificia Universidad Catolica de Chile, Santiago, Chile, 2Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Santiago, Chile, 3Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, 4Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 5Department of Radiology, School of Medicine, Pontificia Universidad Catolica de Chile, Santiago, Chile |
||
Compared to Quantitative Susceptibility Mapping (QSM) in the brain, abdominal QSM faces additional issues due to the presence of gas and fatty tissue. Recent works in abdominal QSM are more focused on its feasibility as a biomarker for disease diagnosis than improving or assessing the robustness and quality of the reconstructions. In this work, we present an abdominal QSM phantom with realistic tissue textures. Our flexible simulation pipeline allows emulating different stages of diseases and MRI signal contributions. Our reconstruction experiments show the potential of our phantom to compare QSM algorithms in different scenarios. |
| 1986 | Computer 81
|
A CNN for Oxygen Extraction Fraction Mapping with combined QSM and qBOLD |
| Patrick Kinz1 and Lothar Schad1 | ||
1Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany |
||
MRI-based mapping of oxygen extraction fraction with QSM and qBOLD is a non-invasive diagnostic tool with many possible applications. But current reconstruction methods based on quasi-Newton (QN) methods are very dependent on accurate parameter initialization. Artificial Neural Networks showed a lot of potential in our previous works. Using a Convolutional Neural Network improves the reconstruction, since neighboring voxels can provide additional information. Using a GESFIDE sequence to sample the qBOLD signal instead of a standard mGRE that samples only the FID, improves the reconstruction accuracy of R2, Y and χnb a lot. |
| 1987 | Computer 82
|
Development of an image reconstruction algorithm to assess brain oxygenation in neonatal asphyxia using quantitative susceptibility mapping |
| Daniel Ridani1,2, Noée Ducros-Chabot1,2, Mathieu Dehaes2,3, and Benjamin De-Leener1,2,4 | ||
1NeuroPoly Lab, Institute of Biomedical Engineering, Polytechnique Montréal, Montreal, QC, Canada, 2Research Center, Ste-Justine Hospital University Centre, Montreal, QC, Canada, 3Department of Radiology, Radio-oncology, and Nuclear Medicine, Université de Montréal, Montreal, QC, Canada, 4Department of Computer and Software Engineering, Polytechnique Montréal, Montreal, QC, Canada |
||
Neonatal asphyxia is a condition that causes lack of oxygen in the infant brain which lead to adverse neurodevelopmental outcomes. MRI provides a very clear image of the brain which lead to a detection of a variety of brain abnormalities. Quantitative susceptibility mapping (QSM) is a MRI tool to quantify brain oxygenation. In this study, we developed QSM reconstruction algorithms and tested them in data acquired in patients following neonatal asphyxia. Results include the comparison of five methods to generate QSM images in the brain specific segmented regions. Our preliminary data suggest that image intensity depends on the reconstruction method. |
| 2053 | Computer 43
|
On the concept of long-range short-wave motion tracking |
| Christoph Michael Schildknecht1 and Klaas Paul Prüssmann1 | ||
1Institute for Biomedical Engineering, ETH Zurich and University of Zurich, Zürich, Switzerland |
||
Current motion tracking modalities operate either in frequency bands that are also occupied by the MRI scanner or interact heavily with the environment. As an alternative, the frequency band of 1-5 MHz can be utilized for motion tracking and has been demonstrated to deliver high performance. Due to the limited sensitivity at a great distance of current broadband systems, we show in this work how short-wave motion tracking can be done at long ranges, enabling independent and robust motion tracking with fewer geometric constraints and greater ease of use. |
| 2054 | Computer 44
|
ROI-focused Non-Rigid Groupwise Registration approach for Motion Correction in ASL Renal Blood Flow Imaging |
| Anne Oyarzun1, Rebeca Echeverria-Chasco2,3, Paloma L. Martin Moreno4, Nuria Garcia-Fernandez4, Gorka Bastarrika2,3, María A. Fernández-Seara1,3, and Arantxa Villanueva1,3,5 | ||
1Electrical, Electronics and Communications Engineering Department, Universidad Pública de Navarra, Pamplona, Spain, 2Radiology, Clínica Universidad de Navarra, Pamplona, Spain, 3IdiSNA, Instituto de Investigación Sanitaria de Navarra, Pamplona, Spain, 4Nephrology, Clínica Universidad de Navarra, Pamplona, Spain, 5ISC, Instituto de Smart Cities, Pamplona, Spain |
||
Motion correction methods are a prerequisite in multiple-image registration tasks. We implemented a non-rigid groupwise registration method and ROI-focused non-rigid groupwise registration method for renal pCASL images. We evaluated the temporal signal variation in the renal cortex after motion correction using both methods. Motion correction technique shows statistically significant improvement on the tSNR and ROI-focused method performs statistically better. |
| 2055 | Computer 45
|
A deep learning image based calibration model to predict motion using auxiliary sensors |
| Radhika Tibrewala1, Mahesh B Keerthivasan2, Kai Tobias Block1, Jan Paska1, and Ryan Brown1 | ||
1CAI2R, NYU School of Medicine, New York, NY, United States, 2Siemens Medical Solutions USA, New York, NY, United States |
||
For MRI hindered by motion artifacts sensors are able to provide a surrogate for bulk motion, but may not be tissue specific. In this work, we use deep learning to build a motion model with an auxiliary pilot tone sensor during a fast-image calibration step. A neural network is used to learn the correlation between the pilot tone signal and the pixel-wise liver displacement which is predicted using an automatic segmentation model. The motion model is used to predict displacement on low frame rate images, thus offering the opportunity to perform motion resolved reconstruction. |
| 2056 | Computer 46
|
Virtual coil concept applied to motion corrected and low rank undersampled reconstructions |
| Gastao Cruz1, Camila Munoz1, René M. Botnar1, and Claudia Prieto1 | ||
1King's College London, London, United Kingdom |
||
Partial Fourier strategies leverage the Hermitian symmetric properties of k-space into higher acceleration factors. Motion corrected (MC) reconstructions incorporate generalized motion into the process. Contrast resolved reconstructions, e.g. low rank inversion (LRI), exploit temporal redundancies in highly accelerated acquisitions. Both the MC and LRI reconstructions are usually ill-posed and can benefit from additional sources of encoding. The Virtual Coil Concept (VCC) has been recently introduced, combining Parallel Imaging with Partial Fourier. Here we investigate the combination of VCC with MC and VCC with LRI to enable further acceleration of both approaches. |
| 2057 | Computer 47
|
Magnetic Resonance Thermometry Motion Compensation during Focused Ultrasound Controlled Hyperthermia in a Small Animal Model |
| Suzanne Wong1,2, Claire Wunker3,4, Ben Keunen2, Maryam Siddiqui5, Karolina Piorkowska2, Yael Babichev4, Warren Foltz6, Rebecca Gladdy3,4, Samuel Pichardo5, Adam Waspe2, and James Drake1,2 | ||
1Biomedical Engineering, University of Toronto, Toronto, ON, Canada, 2Neuroscience and Mental Health, The Hospital for Sick Children, Toronto, ON, Canada, 3Institute of Medical Sciences, University of Toronto, Toronto, ON, Canada, 4Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, ON, Canada, 5Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada, 6Radiation Physics, University Health Network, Toronto, ON, Canada |
||
Magnetic resonance guided high intensity focused ultrasound (MRgHIFU) has gained interest over the past decade due to its ability to administer controlled hyperthermia for localized drug release. One of the main challenges is that MR thermometry is highly susceptible to motion artifacts. A hybrid principal component analysis and projection onto dipole fields motion artifact removal method was applied in real-time during controlled hyperthermia in a murine model using a small-animal MRgHIFU system (Bruker 7T MRI and IGT HIFU). For a target temperature of 40.5°C to be maintained, a significant increase in ultrasound power was required when tissue motion was observed. |
| 2058 | Computer 48
|
Novel methods for correcting motion regression errors caused by global intensity changes in scans of cerebrovascular function |
| Ryan Beckerleg1, Joseph Whittaker1,2, Daniel Gallichan3, and Kevin Murphy1 | ||
1Cardiff University Brain Research Imaging Centre (CUBRIC), School of Physics and Astronomy, Cardiff University, Cardiff, United Kingdom, 2Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 3Cardiff University Brain Research Imaging Centre (CUBRIC), School of Engineering, Cardiff University, Cardiff, United Kingdom |
||
Previously we showed that when global intensity changes (GIC’s) are present in data (e.g., CO2 stimuli during measurement of CVR, ASL tagging), volume registration algorithms misrepresent such signal change as motion. Here we look at motion estimates derived from VRA’s, an external marker-less camera, and novel data-derived techniques, to evaluate if this problem can be overcome. We determined that in most cases where GIC’s are present, the use of an ICA in correction can improve erroneous results from VRA estimates. This isn’t the case in all scans however and the GIC causing this error should be removed prior to correction. |
| 2059 | Computer 49
|
Prospective Motion Correction for a Full Neuroimaging Protocol in Elderly Subjects at 7T |
| Mackenzie Carlson1, Phillip DiGiacomo1, Brian Burns2, Pascal Spincemaille3, Yi Wang3, Julian Maclaren1,4, Murat Aksoy4, Brian Rutt1, and Michael Zeineh1 | ||
1Stanford University, Stanford, CA, United States, 2GE Healthcare, Stanford, CA, United States, 3Weill Cornell Medicine, New York City, NY, United States, 4Hobbitview Inc., San Jose, CA, United States |
||
Correction of head motion during MRI scanning is critical to achieving high-resolution images to study and quantify small regions such as the hippocampus in Alzheimer’s disease. Our optical prospective motion correction prototype system has previously demonstrated improved image quality in volunteers at 7T, but clinical translatability has been lacking. In this work, we implemented an integrated power supply and applied prospective correction to a broad set of sequences, including mGRE and FSE, for a clinically relevant research protocol with emphasis on hippocampal iron imaging. Metrics achievable using our protocol include high-resolution cortical QSM, hippocampal subfield QSM and volume. |
| 2060 | Computer 50
|
MoCoPad: A new soft sensor system for fast head motion detection and tracking in MRI |
| Saikat Sengupta1,2, Mishek Musa3, and Yue Chen4 | ||
1Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, 2Department of Radiology, Vanderbilt University Medical Center, Nashville, TN, United States, 3Department of Mechanical Engineering, University of Arkansas, Fayetteville, AR, United States, 4Department of Biomedical Engineering, Georgia Institute of Technology, Atlanta, GA, United States |
||
In this abstract we present initial feasibility results of a novel pneumatic sensor-based head motion detection and tracking system for MRI. The system comprises of head pad with a built in matrix of air pressure sensors, which replaces the standard head pad in an MRI scan and allows fast sequence agnostic tracking of head motions without the need for navigators, head markers or camera systems. Here, we show initial results for motion detection and head pose estimation in phantom studies using a motion model trained outside the scanner. |
| 2061 | Computer 51
|
Evaluating the match of image quality metrics with radiological assessment in a dataset with and without motion artifacts |
| Hannah Eichhorn1, Simon Chemnitz-Thomsen1,2, Evangelos Vouros1,3, Nitesh Shekhrajka4, Robert Frost5,6, André van der Kouwe5,7, and Melanie Ganz1,8 | ||
1Neurobiology Research Unit, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark, 2Department of Mathematical Sciences, University of Copenhagen, Copenhagen, Denmark, 3Niels Bohr Institute, University of Copenhagen, Copenhagen, Denmark, 4Department of Radiology, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark, 5Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 6Department of Radiology, Harvard Medial School, Boston, MA, United States, 7Department of Radiology, Harvard Medical School, Boston, MA, United States, 8Department of Computer Science, University of Copenhagen, Copenhagen, Denmark |
||
Currently no image quality metric, used for evaluating the performance of artifact correction or image reconstruction methods, is sensitive to all possible image artifacts. This complicates the choice of a proper quantitative quality measure. To provide assistance with this choice, we investigated the correlation of commonly used metrics with radiological evaluation on a dataset acquired with and without motion. The full-reference metrics SSIM and PSNR correlated most strongly with observer scores. Among the reference-free metrics Image Entropy, Average Edge Strength and Tenengrad measure showed a consistent correlation for all investigated sequences. |
| 2062 | Computer 52
|
Choreography Controlled (ChoCo) by synchronized video for motion artifact reproducibility and dataset generation |
| Oscar Albert Dabrowski1, Sébastien Courvoisier2, Jean-Luc Falcone3, Antoine Klauser2, Julien Songeon3, Michel Kocher2, Bastien Chopard3, and Francois Lazeyras2 | ||
1Computer science, University of Geneva, Geneva, Switzerland, 2CIBM, Geneva, Switzerland, 3University of Geneva, Geneva, Switzerland |
||
In the age of artificial intelligence, there is a need for simple and inexpensive frameworks aimed at performing standardized reproducible motion-controlled experiments allowing the creation of datasets of motion corrupted in vivo MRI scans. Focused on human brain imaging, we propose ChoCo: a choreography controlled protocol for reproducible head motion and an ad-hoc hardware device for synchronization between an MRI scanner and a movie displaying motion to be performed. A proof of concept demonstrated that 6 participants were able to reproduce head choreographies accurately. The resulting motion corrupted brain images show qualitatively similar artifacts, confirming consistent motion reproduction among subjects. |
| 2063 | Computer 53
|
Rotation estimation for cloverleaf navigators using a k-space map simulation |
| Tamsin Edwards Lambourne1,2, André J. W. van der Kouwe1,3, and Robert Frost1,3 | ||
1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, United States, 2Northeastern University, Boston, MA, United States, 3Department of Radiology, Harvard Medical School, Boston, MA, United States |
||
Short <5 ms cloverleaf navigators have been demonstrated for prospective motion correction at a rate of 71 Hz in steady-state FLASH. The original cloverleaf approach required a 12 second k-space mapping pre-scan to reliably deal with out-of-plane rotations. Here we investigate simulation of a k-space map based on a low spatial resolution pre-scan. The approach shows potential for real-time detection of rigid-body rotation in the context of real-time self-correction of the cloverleaf orientation. |
| 2064
|
Computer 54
|
3D MR spirometry sensitivity to gravity lung dependence in healthy volunteers |
| Nathalie Barrau1, Claire Pellot Barakat1, Zhongzheng He1, Tolga Emre1, Angéline Nemeth1, Anass Afkir1, Wenpei Cai1, Vincent Brulon1, Tanguy Boucneau2, Brice Fernandez2, Florent Besson1, Vincent Lebon1, and Xavier Maître1 | ||
1Université Paris-Saclay, CEA, CNRS, Inserm, BioMaps, Orsay, France, 2General Electric Healthcare, Buc, France |
||
Spirometry is a standard exam in pulmonary function testing for assessing the efficiency of global ventilation from flow-volume loops measured at the mouth during forced respiratory cycles. A newly developed technique, 3D MR spirometry, makes use of dynamic magnetic resonance imaging during spontaneous breathing to produce regional characterization of the lung function. The sensitivity of the technique was challenged with respect to gravity by extracting the associated ventilation gradients in prone and supine healthy volunteers. |
| 2065 | Computer 55
|
3D super-resolution MR elastography of the brain |
| Simone Hufnagel1, Matthias Anders2, Christoph Stefan Aigner1, Helge Herthum2, Heiko Tzschaetzsch2, Tobias Schaeffter1,3,4, Ingolf Sack2, and Christoph Kolbitsch1 | ||
1Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany, 2Radiology, Charité - Universitätsmedizin Berlin, Berlin, Germany, 3School of Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom, 4Department of Biomedical Engineering, Technical University of Berlin, Berlin, Germany |
||
MR elastography (MRE) provides valuable quantitative information about the mechanical properties of brain tissues. However, due to SNR limitations, often only low through-plane resolution is possible. We present super-resolution MRE based on multiple stacks of complex 3D wavefields of the brain resulting in elastograms with isotropic (1×1×1) mm3 resolution. The approach was evaluated for the in-vivo brain and showed improved visibility of fine structures while presenting consistent shear wave speed values. |
| 2066 | Computer 56
|
Local Integral EPT - the alter ego of Helmholtz EPT |
| Luca Zilberti1, Alessandro Arduino1, Umberto Zanovello1, and Oriano Bottauscio1 | ||
1INRIM, Torino, Italy |
||
A local-integral solution of the Electric Properties Tomography (EPT) problem is proposed. The method exploits a trick to consider the Helmholtz equation that regulates EPT as a Poisson equation. Then, it inverts the equation under the local homogeneity assumption, using Green's theorem. The approach turns out to be equivalent to the traditional Helmholtz-EPT. |
| 2067 | Computer 57
|
GluCEST in Deep Gray Matter and Adjoining White Matter of pediatric-onset Multiple Sclerosis Brains using a Robust B1-calibration Approach |
| Dushyant Kumar1, Ritobrato Datta2, Micky K Bacchus2, Narayan Datt Soni1, Ravi Prakash Reddy Nanga1, Brenda Banwell 2, and Ravinder Reddy1 | ||
1Radiology, Center for Advanced Metabolic Imaging in Precision Medicine (CAMIPM), University of Pennsylvania, Philadelphia, PA, United States, 2Neurology, Children's Hospital of Philadelphia, Philadelphia, PA, United States |
||
Several proton magnetic resonance spectroscopy (1H-MRS) based studies have reported regional alterations of glutamate metabolism in both acute and chronic multiple sclerosis (MS) pathologies, including in normal appearing white matter and mixed tissues. Despite its well-established capability for detecting altered glutamate metabolism, 1H-MRS is hampered by the low spatial resolution that precludes the measurements from small MS lesions. We demonstrate the feasibility of performing glutamate weighted imaging using chemical exchange saturation transfer with B0- and B1- corrections in pediatric-onset MS subjects. The proposed model for B1-calibration is a major improvement over the phenomenological method previously proposed by our group. |
| 2068 | Computer 58
|
Quantifying myelin water exchange using optimized bSSFP sequences |
| Naveen Murthy1, Jon-Fredrik Nielsen2, Steven T. Whitaker1, Melissa W. Haskell1,2, Scott D. Swanson3, Nicole Seiberlich3, and Jeffrey A. Fessler1 | ||
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States, 2fMRI Lab, University of Michigan, Ann Arbor, MI, United States, 3Radiology, University of Michigan, Ann Arbor, MI, United States |
||
Mapping myelin water exchange rates could potentially be useful in helping us understand and characterize different diseases in the brain. It is a challenging quantitative parameter mapping problem to estimate multi-compartment exchange. This work focuses on scan design, with an aim of estimating exchange maps with good precision. In particular, a set of balanced steady-state free precession (bSSFP) sequences are optimized to estimate exchange rates in white matter (WM), for a two-pool model. Quantitative exchange maps are shown for a simulated phantom, using this optimized scan design. |
| 2069 | Computer 59
|
MRI-based radiomics quantification for characterization of clear cell renal cell carcinoma. |
| Octavia Bane1,2, Christopher Kyriakakos1, Nicolas Gillingham1, Jordan Cuevas1,2, Kirolos Meilika3, Jorge Daza3, Amir Horowitz4, Bachir Taouli1,2, Ketan Badani3, and Sara Lewis1,2 | ||
1Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 3Department of Urology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 4Division of Oncological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, United States |
||
In this prospective, single-center study, we evaluated the use of qualitative features and quantitative radiomics features from MRI for diagnosis and characterization of clear cell renal cell carcinoma (ccRCC). Qualitative features in isolation were of no value for diagnosing ccRCC, while the ccRCC likelihood score, which takes into account qualitative T2w signal and contrast enhancement pattern, had comparable diagnostic performance to the quantitative radiomics features. Quantitative radiomics features had good to excellent diagnostic performance in identifying ccRCC, as well as correlated to grade of ccRCC. |
| 2070 | Computer 60
|
Using MRI-based fat fraction histograms in dystrophic muscle to assess individual patient disease evolution |
| Harmen Reyngoudt1,2, Pierre-Yves Baudin1,2, Ericky Caldas de Almeida Araujo1,2, Brenda L Wong3,4, Pierre G Carlier5, and Benjamin Marty1,2 | ||
1NMR Laboratory, Neuromuscular Investigation Center, Institute of Myology, Paris, France, 2NMR Laboratory, CEA/DRF/IBFJ/MIRCen, Paris, France, 3Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, United States, 4UMass Memorial Medical Center, Worcester, MA, United States, 5University Paris-Saclay, CEA, EA, DRF, Service Hospitalier Frédéric Joliot, Orsay, France |
||
In most clinical quantitative MRI studies, using fat-water imaging, a mean fat fraction value per region of interest is generally used. The aim of this study was to look into the longitudinal changes in FF distribution on an individual patient basis, in dystrophic muscle, and at the same time, investigate whether differences were observed between treated and non-treated patients. As shown here, even if the mean fat fraction is the same between two patients or across time, there might be significant differences in the respective FF distributions, and might reveal, in retrospect, differential clinical or functional changes. |
| 2071 | Computer 61
|
Data-driven B1 non-uniformity correction in ultra-high-field small-animal macromolecular proton fraction and R1 mapping |
| Vasily L. Yarnykh1, Alena A. Kisel1, Andrey E. Akulov2, and Alexander Drobyshevsky3 | ||
1Radiology, University of Washington, Seattle, WA, United States, 2Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk, Russian Federation, 3NorthShore University HealthSystem, Evanston, IL, United States |
||
The data-driven surrogate B1 field correction algorithm was recently developed to eliminate B1-related spatial errors in the macromolecular proton fraction (MPF) and R1 maps obtained using the fast single-point method. The algorithm obviates the need for acquisition of actual B1 field maps, thus extending the routine use of fast MPF mapping. Initially, the surrogate B1 correction was applied to the human brain imaging at 3T. In this study, the method has been adopted and validated for small-animal imaging utilizing dedicated ultra-high-field MRI scanners. Particularly, we report guidelines on using surrogate B1 field correction in 9.4T and 11.7T magnetic fields. |
| 2072 | Computer 62
|
Towards validation of in vivo U-fibre mapping using post-mortem DWI tractography of V1-V2 connections |
| Fakhereh Movahedian Attar1, Evgeniya Kirilina 1,2, Christian Schneider1,3, Daniel Haenelt1, Luke J. Edwards1, Kerrin J. Pine1, Carsten Jäger1,4, Katja Reimann1, Andreas Pohlmann5, Joao Periquito5,6, Tobias Streubel7, Siawoosh Mohammadi1,7, Thoralf Niendorf5,8, Markus Morawski4, and Nikolaus Weiskopf1,9 | ||
1Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Department of Education and Psychology, Center for Cognitive Neuroscience Berlin, Free University Berlin, Berlin, Germany, 3Faculty of Physics and Earth Sciences, Leipzig University, Leipzig, Germany, 4Paul Flechsig Institute of Brain Research, University of Leipzig, Leipzig, Germany, 5Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany, 6Institute of Physiology, Charité- Universitätsmedizin Berlin, Berlin, Germany, 7Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, 8Experimental and Clinical Research Center, a joint cooperation between the Charité Medical Faculty and the Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany, 9Felix Bloch Institute for Solid State Physics, Faculty of Physics and Earth Sciences, Leipzig University, Leipzig, Germany |
||
U-fibres are the most abundant white matter fibres and yet are highly underrepresented in the MRI-derived human brain connectome. Development of validated in vivo pipelines for comprehensive mapping of these short association fibres is therefore essential. Here, we show the correspondence of the geometries of U-fibres connecting early visual cortices mapped in vivo using sub-millimetre resolution diffusion weighted imaging (DWI) tractography to those obtained in a post-mortem brain tissue sample using ultra-high resolution DWI tractography. |
| 2154 | Computer 56
|
Incorporating Histological Grading for Brain Tumor Segmentation |
| Lipei Zhang1, Yiran We2, Chao Li1,2, Stephen John Price2, and Carola Bibiane Schönlieb1 | ||
1The Centre of Mathematical Imaging in Healthcare, Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge, United Kingdom, 2Cambridge Brain Tumor Imaging Laboratory, Division of Neurosurgery, Deoartment of Clinic Neuroscience, Univerisity of Cambridge, Cambridge, United Kingdom |
||
Automatic tumor segmentation on MRI is crucial for patient management of brain tumors. However, previous automatic MRI segmentation methods based on deep learning are limited by bottlenecks of feature extraction. In this paper, we seek to investigate the influence of incorporating prior histology grading on the performance of tumor segmentation. We propose to employ histological grade labels to strengthen the feature extraction and tumor localization in the encoder, which can boost the DSC at exceeding 2% in different U-Net baselines. Our results could support the usefulness of incorporating clinical prior knowledge to improve the segmentation performance without introducing extra parameters. |
| 2155 | Computer 57
|
Automated brain morphometry for sub-millimeter 7T MRI using transfer learning |
| Gian Franco Piredda1,2,3, Punith B. Venkategowda4, Piotr Radojewski5,6, Tom Hilbert1,2,3, Arun Joseph6,7,8, Gabriele Bonanno6,7,8, Roland Wiest5,6, Karl Egger9, Shan Yang9, Jean-Philippe Thiran2,3, Ricardo A. Corredor-Jerez1,2,3, Bénédicte Maréchal1,2,3, and Tobias Kober1,2,3 | ||
1Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, 2Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 3LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, 4Siemens Healthcare Pvt. Ltd., Bangalore, India, 5Support Center for Advanced Neuroimaging, Institute for Diagnostic and Interventional Neuroradiology, Inselspital, Bern University, Bern, Switzerland, 6Translational Imaging Center, sitem-insel AG, Bern, Switzerland, 7Advanced Clinical Imaging Technology, Siemens Healthcare AG, Bern, Switzerland, 8Magnetic Resonance Methodology, Institute of Diagnostic and Interventional Neuroradiology, University of Bern, Bern, Switzerland, 9Department of Neuroradiology, Medical Center – University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany |
||
Large spatial signal variations due to field inhomogeneities complicate the application of automated brain morphometry at 7T. In this work, we propose to use transfer learning to adapt a template-based segmentation algorithm to sub-millimeter ultra-high field applications. More specifically, a convolutional neural network pre-trained on T1-weighted scans to extract the total intracranial volume (TIV) from MP-RAGE acquisitions was re-trained to retrieve the TIV mask directly from MP2RAGE volumes. The developed method proved to reliably deliver brain tissue masks and volumetry at 7T. |
| 2156 | Computer 58
|
A privacy-preserving federated learning infrastructure for prostate segmentation on T2-Weighted MRI |
| Fadila Zerka1, Mohammed Sunoqrot1, Bendik Abrahamsen1, Alexandros Patsanis1, Tone Frost Bathen1,2, and Mattijs Elschot1,2 | ||
1Department of Circulation and Medical Imaging, NTNU, Trondheim, Norway, 2Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim, Norway |
||
Accessing medical data is highly protected by law and ethics, making data sharing difficult and time-consuming. Distributed learning in its various forms allows learning from medical data without these data ever leaving the medical institutions. In this study, we evaluate the Flower federated learning framework for prostate segmentation on T2-Weighted MRI. The results show that the Federated learning framework performs comparably to the reference (centralized learning) model. |
| 2157 | Computer 59
|
Reducing the impact of texture on deep-learning brain tissue segmentation networks trained with simulated MR images |
| Yasmina Al Khalil1, Aymen Ayaz1, Cristian Lorenz2, Jürgen Weese2, Josien Pluim1, and Marcel Breeuwer1,3 | ||
1Biomedical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands, 2Philips Research Laboratories, Hamburg, Germany, 3Philips Healthcare, MR R&D - Clinical Science, Best, Netherlands |
||
Deep learning-based segmentation algorithms largely rely on the availability of extensive, clinically representative data. While collecting such data requires vast resources, image simulation has the potential of generating realistic data reproducing a wide range of scanner sequences or parameters. In this work, we present an MR image simulation pipeline and evaluate its potential for training a deep-learning network for segmenting several brain structures in T1-weighted images acquired from real scanners. We additionally demonstrate how to prevent performance degradation from the lack of tissue texture in simulated images by combining statistical texture analysis and filtering on the evaluation image set. |
| 2158 | Computer 60
|
Efficient, streamline-free white matter tract segmentation for intraoperative diffusion MRI. |
| Fiona Young1, Patrick Hales1,2, Laura Mancini3,4, Sotirios Bisdas3,4, Caroline Micallef3,4, Tarek Yousry3,4, Christopher A. Clark1, Kristian Aquilina5, and Jonathan D. Clayden1 | ||
1UCL Great Ormond Street Institute of Child Health, University College London, London, United Kingdom, 2Department of Radiology, Great Ormond Street Hospital for Children, London, United Kingdom, 3Lysholm Department of Neuroradiology, The National Hospital for Neurology & Neurosurgery, University College London Hospitals NHS Foundation Trust, London, United Kingdom, 4Neuroradiological Academic Unit, Department of Brain Repair and Rehabilitation, UCL Queen Square Institute of Neurology, London, United Kingdom, 5Department of Neurosurgery, Great Ormond Street Hospital for Children, London, United Kingdom |
||
Tractfinder is an atlas-based approach to segmenting white matter tracts without tractography, in subjects with space-occupying lesions. A shape and orientation prior is combined with fibre orientation distributions in the target data to produce a map of tract location. Results for a given tract can be obtained in a matter of minutes and are comparable to those afforded by clinical tractography. In future, tractfinder could be applied to intraoperative diffusion MRI to aid neuronavigation during brain surgery. |
| 2159 | Computer 61
|
Infant Brain Segmentation based on Multi-dMRI contrast Attention (BISMA) |
| Tianjia Zhu1,2, Minhui Ouyang1, Lei Feng3, Kay Sindabizera1, Jessica Hyland1, and Hao Huang1,4 | ||
1Department of Radiology, Children's Hospital of Philadelphia, Philadelphia, PA, United States, 2Department of Bioengineering, University of Pennsylvania, Philadelphia, PA, United States, 3Shandong University Cheeloo College of Medicine, Jinan, China, 4Department of Radiology, University of Pennsylvania, Philadelphia, PA, United States |
||
On widely used relaxation-based T1-weighted (T1w) and T2-weighted (T2w) MRI images, 0-2-year-old infant brain images exhibit poor gray and white matter contrasts due to dynamic and poor myelination in this stage. Such poor T1w and T2w contrasts hinder accurate segmentation of infant brains based on T1w or T2w images. Instead, diffusion MRI (dMRI)-derived maps offer high contrasts throughout infancy. To leverage rich dMRI contrasts, we established an Infant Brain Segmentation based on Multi-dMRI contrast Attention (BISMA) technique. BISMA fills the gap in accurate deep learning segmentation of infant brain by fusing multi-contrasts of dMRI-derived maps. |
| 2160 | Computer 62
|
Generalisability of Automated CNN-based Renal Segmentation for Multi-Vendor Studies |
| Alexander J Daniel1, Charlotte E Buchanan1, David M Morris2, Hao Li3, Rebecca Noble1, João Sousa4, Steven Sourbron4, David L Thomas5,6,7, Andrew N Priest3,8, and Susan T Francis1 | ||
1Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, United Kingdom, 2Centre for Cardiovascular Science, University of Edinburgh, Edinburgh, United Kingdom, 3Department of Radiology, University of Cambridge, Cambridge, United Kingdom, 4Department of Infection, Immunity and Cardiovascular Disease, University of Sheffield, Sheffield, United Kingdom, 5Neuroradiological Academic Unit, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 6Dementia Research Centre, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 7Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 8Department of Radiology, Addenbrooke’s Hospital, Cambridge, United Kingdom |
||
Manual segmentation of the kidneys is very time consuming and reader dependent, this renders measurements of total kidney volume (TKV) in large multi-site studies impractical. Here we use a convolutional neural network (CNN), trained on data from a single MRI vendor, to segment the kidneys of volunteers scanned with a harmonised FSE image protocol on MR scanners from three different vendors (GE, Philips and Siemens). The kidneys were manually segmented by two readers, both of which demonstrated a significant difference in TKV across vendors; no significant difference in TKV was found in the segmentations produced by the CNN. |
| 2161 | Computer 63
|
High temporal-resolution MRI during mild-cold exposure enables the assessment of brown adipose tissue with a low inter-image variability |
| Aashley S.D. Sardjoe Mishre1,2, Maaike E. Straat2, Borja Martinez-Tellez2, Mariëtte R. Boon2, Oleh Dzyubachyk3,4, Andrew G. Webb1, Patrick C.N. Rensen2, and Hermien E. Kan1 | ||
1Department of Radiology, C.J. Gorter Center for High Field MRI, Leiden University Medical Center, Leiden, Netherlands, 2Department of Medicine, Division of Endocrinology and Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden, Netherlands, 3Department of Radiology, Division of Image Processing (LKEB), Leiden University Medical Center, Leiden, Netherlands, 4Department of Cell and Chemical Biology, Electron Microscopy section, Leiden University Medical Center, Leiden, Netherlands |
||
Brown adipose tissue (BAT) is considered to be a potential therapeutic target against cardiometabolic diseases. Activated BAT combusts intracellular fatty acids leading to a reduction in fat fraction . Both cold exposure and pharmacological stimuli can activate BAT, but the short-term dynamics of BAT activation are unknown. To assess supraclavicular BAT fat fraction dynamics during cold-exposure, we developed a 1-minute-time-resolution MRI protocol using breath-holds and co-registration to minimize motion-artefacts. We demonstrated the validity and feasibility of our image analysis method, and found an inter-image variability of less than 0.1% fat fraction. |
| 2162 | Computer 64
|
Automatic segmentation of whole-body MRI using UnnU-Net: Feasibility of whole-skeleton ADC evaluation in plasma cell disorders |
| Renyang Gu1, Michela Antonelli2, Pritesh Mehta 3, Ashik Amlani 4, Adrian Green4, Radhouene Neji 5, Sebastien Ourselin2, Isabel Dregely2, and Vicky Goh2 | ||
1Department of Biomedical Engineering and Imaging Science, King's College London, London, United Kingdom, 2School of Biomedical Engineering & Imaging Sciences, King's College London, London, United Kingdom, 3University College London, London, United Kingdom, 4Department of Radiology, Guy’s & St Thomas’ NHS Foundation Trust, London, United Kingdom, 5Siemens Healthcare Limited, London, United Kingdom |
||
Multiple myeloma is a heterogeneous bone marrow cancer. Assessment of changes in mean apparent diffusion coefficient (ADCmean) is helpful for evaluating treatment response but has not been feasible for the whole skeleton due to the time-consuming nature of manual segmentation. Whole-skeleton and per-station ADCmean were quantified from whole-body MRI using automated segmentation by an uncertainty-aware nnU-Net in 30 patients with plasma cell disorders and compared against the manual segmentation by radiologists. No differences were observed in whole-skeleton or per-station ADCmean when using the automatic and manual segmentations. Further investigation is required in a larger dataset, but initial results are promising. |
| 2163 | Computer 65
|
Protocol adaptive Stacked transfer learning (STL) U-NET with small dataset training for soft tissue segmentation in dynamic speech MRI |
| Subin Erattakulangara1, Karthika Kelat1, and Sajan Goud Lingala1,2 | ||
1Roy J Carver Department of Biomedical Engineering, University of Iowa, Iowa city, IA, United States, 2Department of Radiology, University of Iowa, Iowa city, IA, United States |
||
We develop a protocol adaptive Stacked transfer learning (STL) U-NET for soft tissue segmentation in dynamic speech MRI. Our approach leverages knowledge from large open-source datasets, and only needs to be trained on small number of protocol specific images (of the order of 20 images). We demonstrate the utility of STL U-NET in efficiently segmenting soft-tissue articulators from three different protocols with different field strengths, vendors, acquisition, reconstruction. Using the DICE similarity metric, we demonstrate segmentation accuracies with our approach to be at the level of manual segmentation. |
| 2164 | Computer 66
|
A deep-learning non-contrast MRI lesion segmentation model for liver cancer patients that may have had previous surgery |
| Nora Vogt1, Zobair Arya1, Luis Núñez1, John Connell1, and Paul Aljabar1 | ||
1Perspectum, Oxford, United Kingdom |
||
Segmentation of lesions within the liver is pivotal for surveillance, diagnosis and treatment planning, and automated or semi-automated approaches can aid clinical workflows. Many patients that are under surveillance may have had previous surgery, meaning their scans will contain post-surgical features. We investigate the use of a deep learning segmentation model for non-contrast MRI that can distinguish between lesions, surgical clips and resection-induced fluid-filling regions. We report mean dice scores of 0.55, 0.72 and 0.88 for these classes respectively, demonstrating the potential of this model for semi-automated workflows. |
| 2223 | Computer 36
|
Lung segmentation with deep learning for 3D MR spirometry |
| Zhongzheng He1,2, Nathalie Barrau1, Claire Pellot Barakat1, and Xavier Maître1 | ||
1Université Paris-Saclay, CEA, CNRS, Inserm, BioMaps, Orsay, France, 2IADI U1254, INSERM, Université de Lorraine, Nancy, France |
||
Current MR lung image segmentation has huge challenges compared to CT images, specifically in terms of low contrast, non-homogeneity. We developed a series of processing to accelerate the manual correction of reference standard lung masks by the auto-seeds region growing. Finally, we developed a 3D automatic MRI lung segmentation method using deep learning with a limited dataset (41 volumes for training and 4 volumes for validation). In the primary result, this lung segmentation archived a Dice score of 0.917±0.013. In the case of limited data, it provides us a new way for MR lung segmentation. |
| 2224 | Computer 37
|
Hippocampal Subfields Volume in Middle Age Healthy Adults |
| salem Alkhateeb1, Tales Santini1, Li jinghang1, Robin Chu1, Daniel Ibrahim1, Anna M. Marsland1, Stephen B. Manuck1, Pete Gianaros1, and tamer S. Ibrahim2 | ||
1University of Pittsburgh, Pittsburgh, PA, United States, 2University of Pittsburgh, pittsburgh, PA, United States |
||
As hippocampal volume has been extensively utilized as a diagnosing tool to confirm diagnosis of many neurological disorders, this study aims to employ the high resolution 7T TSE T2w data to segment high quality images with precision based on multi atlases and machine learning. Results have proved that this pipeline provides excellent outcomes and is validated to be used for more variables. |
| 2225 | Computer 38
|
Effect of Compressed SENSE on Freesurfer parcellation precision |
| Michael A Green1,2, Peter Humberg3, Iain K Ball4, and Caroline D Rae1,2 | ||
1Neuroscience Research Australia, Sydney, Australia, 2School of Medical Sciences, University of New South Wales, Sydney, Australia, 3Stats Central, Mark Wainwright Analytical Centre, University of New South Wales, Sydney, Australia, 4Philips Australia & New Zealand, Sydney, Australia |
||
We employed the commonly used software package, Freesurfer to obtain volume, area and thickness measurements from structural MR imaging data acquired using the Compressed SENSE acceleration technique. We assessed measurement reliability via equivalence testing on a series of increasing acceleration factors to provide guidelines for researchers wanting to reduce scan acquisition time while acquiring high-quality data. |
| 2226 | Computer 39
|
Automatic fibroglandular tissue segmentation in breast MRI using a deep learning approach |
| Fares Ouadahi1, Anais Bernard1, Lucile Brun1, and Julien Rouyer1 | ||
1Department of Research & Innovation, Olea Medical, La Ciotat, France |
||
An accurate fibroglandular tissue (FGT) segmentation model was designed using of a deep learning strategy on T1w series without fat suppression. The proposed method combined a dedicated preprocessing and the training of a two-dimensional U-Net architecture on a multi-centric representative database to achieve an automatic FGT segmentation. The final test of the generated model exhibited overall good performances with a median Dice similarity coefficient of 0.951. More contrasted performances were obtained when correlating the gland density with the discrepancy between ground truth and prediction. Indeed, the lower the breast density, the greater the uncertainty in the segmentation. |
| 2227 | Computer 40
|
Immersive and Interactive On-the-fly MRI Control and Visualization with a Holographic Augmented Reality Interface: A Trans-Atlantic Test |
| Jose Velazco-Garcia1, Nikolaos Tsekos1, Andrew Webb2, and Kirsten Koolstra2 | ||
1Medical Robotics and Imaging Lab, University of Houston, Houston, TX, United States, 2Leiden University Medical Center, Leiden, Netherlands |
||
We present the computational Framework for Interactive Immersion into Imaging Data (FI3D) for 3D visualization of MR data as they are collected and on-the-fly control of the MRI scanner. The FI3D was implemented to integrate an MRI scanner with the operator who used a customized holographic scene, generated by a HoloLens head mounted display (HMD), to (1) view images and image-generated outcomes (e.g., segmentation) and (2) control the MR scanner to update the acquisition protocol. In this study, the framework linked an MR scanner in Leiden (The Netherlands) and an operator in Houston (USA). |
| 2228 | Computer 41
|
A Framework for Brain Tumor Detection, Classification and Segmentation using Deep Learning |
| Rafia Ahsan1, Iram Shahzadi2,3, Ibtisam Aslam1,4, and Hammad Omer1 | ||
1Medical Image Processing Research Group (MIPRG), Dept. of Elect. & Comp. Engineering, COMSATS University Islamabad, Islamabad, Pakistan, 2OncoRay – National Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Helmholtz-Zentrum Dresden – Rossendorf, Dresden, Germany, 3German Cancer Research Center (DKFZ), Heidelberg, Germany, 4Service of Radiology, Geneva University Hospitals and Faculty of Medicine, University of Geneva, Geneva, Switzerland |
||
Detection, classification and segmentation of brain tumor simultaneously is challenging due to the heterogeneous nature of the tumor. Limited work has been done in literature in this regard. The present study, therefore, aims to identify an object detection network that would be able to solve multi-class brain tumor classification and detection problem with high accuracy. Furthermore, the best performing detection network has been cascaded with 2D U-Net for pixel level segmentation. The proposed method not only classifies the tumor with high accuracy but also provides improved segmentation results compared to the standard U-Net. |
| 2229 | Computer 42
|
Cross-sectional robustness of 6 freely available software packages for brain volume measurements in multiple sclerosis |
| David Rudolf van Nederpelt1, Houshang Amiri1,2, Amanda Mariyampillai1, Iman Brouwer1, Samantha Noteboom3, Frederik Barkhof1,4, Joost P.A. Kuijer1, and Hugo Vrenken1 | ||
1Radiology and nuclear medicine, Amsterdam UMC, location VUmc, Amsterdam, Netherlands, 2Neuroscience research center, Kerman University of Medical Sciences, Kerman, Iran (Islamic Republic of), 3Anatomy and Neurosciences, Amsterdam UMC, location VUmc, Amsterdam, Netherlands, 4Institutes of Neurology and Healthcare Engineering, UCL London, London, United Kingdom |
||
Automated segmentation of brain MR images has paved the way for large cohort atrophy studies in multiple sclerosis (MS). A variety of automated software packages is available. Here we aimed to quantify brain volume differences measured by six freely available software packages on data from 21 MS patients all scanned thrice on three different MR scanners. While intra-class correlation coefficients were high, systematic differences between scanners were found for every software. This suggests that direct comparison of volumes acquired with different scanners is not possible and standardization is needed. |
| 2230 | Computer 43
|
Three-dimensional reconstruction and characterization of bladder deformations |
| Augustin C. Ogier1, Stanislas Rapacchi2, and Marc-Emmanuel Bellemare1 | ||
1Aix Marseille Univ, Universite de Toulon, CNRS, LIS, Marseille, France, 2Aix Marseille Univ, CNRS, CRMBM, Marseille, France |
||
Pelvic floor disorders are prevalent diseases and only 2D dynamic observations of straining exercises at excretion are available in clinics. The understanding of three-dimensional pelvic organs mechanical defects is not yet achievable. We proposed a complete methodology for the 3D representation of the bladder combined with 3D representation of the location of the highest strain areas on the organ surface. We assessed the potential of our method on eight control subjects for the reconstruction of bladder during intense straining from forced breathing exercises. The proposed methodology provided, for the first time, a proper 3D+t spatial tracking of bladder non-reversible deformations. |
| 2231 | Computer 44
|
Deep Learning-Based Automatic Lung Segmentation for MR Images at 0.55T |
| Rachel Chae1,2, Ahsan Javed1, Rajiv Ramasawmy1, Hui Xue1, Marcus Carlsson1, Adrienne E Campbell-Washburn1, and Felicia Seemann1 | ||
1Cardiovascular Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, United States, 2Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States |
||
High-performance 0.55T systems are well suited for structural lung imaging due to reduced susceptibility and prolonged T2*. Lung segmentation is required to derive metrics of pulmonary function from lung MRI. Convolutional neural networks are effective for lung segmentation at higher field strengths, but they do not generalize to images acquired at 0.55T. This study develops a neural network for automated lung segmentation of T1 and proton density weighted ultrashort-TE MRI at 0.55T. Training data was generated using segmentations by active contours and manual corrections. The proposed network was fast (1.07s) and as accurate as existing semi-automated segmentation (dice coefficient 0.93). |
| 2232 | Computer 45
|
Deep learning-based tumor segmentation from postoperative MRI |
| Jingpeng Li1,2, Jonas Vardal3,4, Inge Groote1, and Atle Bjornerud1,2 | ||
1Computational Radiology and Artificial Intelligence, Oslo University Hospital, Oslo, Norway, 2Department of Physics, University of Oslo, Oslo, Norway, 3The Intervention Centre, Oslo University Hospital, Oslo, Norway, 4Faculty of Medicine, University of Oslo, Oslo, Norway |
||
To accurately detect and localize postoperative tumor after surgery is of critical importance to postoperative patient management and survial rate. We propose a fully automated end-to-end coarse-to-fine segmentation approach for the segmentation of posoperative tumor. |
| 2233 | Computer 46
|
Deep learning-based segmentation of the parasagittal dural space from non-contrast anatomical MRI image: a resource for glymphatic studies |
| Kilian Hett1, Colin D. McKnight2, Jarrod J. Eisma1, Jason Elenberger1, Ciaran M. Considine1, Daniel O. Claassen1, and Manus J. Donahue1,2,3 | ||
1Neurology, Vanderbilt University Medical Center, Nashville, TN, United States, 2Radiology and Radiological Sciences, Vanderbilt University Medical Center, Nashville, TN, United States, 3Psychiatry and Behavioral Sciences, Vanderbilt University Medical Center, Nashville, TN, United States |
||
The overarching goal of this work is to develop and validate novel deep learning algorithms for segmenting the parasagittal dural (PSD) space, which has been hypothesized to harbor cerebral lymphatic channels, from standard non-contrast anatomical imaging. Specifically, contrasted based MRI studies have recently suggested that the PSD may be important for CSF clearance, however existing methods for evaluating this space require administration of exogenous contrast and time-consuming manual tracing, thereby limiting generalizability. We propose a new segmentation method using non-contrasted MRI, and we validate this method in a mixed cohort of older adults with and without neurodegeneration. |
| 2234 | Computer 47
|
Split-and-Merge Segmentation in Magnetic Resonance Imaging Based on Graph Wedgelets |
| Wolfgang Erb1 | ||
1Dipartimento di Matematica "Tullio Levi-Civita", University of Padova, Padova, Italy |
||
Graph wedgelets are a novel tool for the fast decomposition of images in geometrically meaningful, wedge-shaped subregions. In this work, we study the usage of graph wedgelets as a promising splitting method in a split-and-merge segmentation scheme for Magnetic Resonance Imaging. We combine adaptive wedgelet splits of MRI images with a simple and classical merging strategy for subregions and obtain in this way an efficient and robust segmentation of diagnostic-relevant subdomains in MRI data. |
| 2235 | Computer 48
|
Cascaded parallel imaging for 3D deep learning reconstruction |
| Jingyuan Lyu1, Yongquan Ye1, Zhongqi Zhang1, Jian Xu1, Xiao Chen2, Terrence Chen2, Shanhui Sun2, and Eric Z. Chen2 | ||
1UIH America, Inc., Houston, TX, United States, 2United Imaging Intelligence, Cambridge, MA, United States |
||
This study demonstrates the effects on image reconstruction by integrating a separate parallel imaging layer with the deep learning module. With such additional layer, deep learning reconstruction is flexible and reliable for highly under-sampled data. |
| 2339 | Computer 75
|
Iterative Model-Based Image Reconstruction of RF gradient-based MRI |
| Taylor Froelich1 and Michael Garwood1 | ||
1Center for Magnetic Resonance Research and Department of Radiology, University of Minnesota, Minneapolis, MN, United States |
||
When reconstructing with traditional Fourier-based techniques, image distortions arising from nonlinear B1 and B0 inhomogeneity can plague radiofrequency-based (RF) imaging methods that rely on B1 gradients for spatial encoding. In this work, we propose a new framework for reconstructing multi-dimensional RF gradient-based images leveraging an iterative approach to solve a regularized inverse problem. The proposed methodology employs a full Bloch simulation to reconstruct an undistorted image after determination of the forward operator and measured receive coil sensitivities. |
| 2340 | Computer 76
|
Intravoxel B0 Corrected Image Reconstruction with RF Prephasing |
| Steven T. Whitaker1, Jon-Fredrik Nielsen2, and Jeffrey A. Fessler1 | ||
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States, 2Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States |
||
Regions with large intravoxel B0 gradients result in a wide spread of off-resonance frequencies within each voxel, causing spins within a voxel to dephase with respect to each other. Using model-based reconstruction to account for this dephasing can help alleviate artifacts from this signal loss, but success is limited in areas of extreme dephasing. We propose a model-based reconstruction method that includes RF prephasing to help mitigate the effects of extreme dephasing. We demonstrate that the proposed approach successfully recovers signal in areas of extreme dephasing and results in lower reconstruction error than model-based reconstruction without RF prephasing. |
| 2341 | Computer 77
|
Reconstruction of 3D EPI timeseries including a correction of different acquisition times |
| Samuel Bianchi1, Jakob Heinzle2, Maria Engel1, Lars Kasper1,2, and Klaas P. Pruessmann1 | ||
1Institute for Biomedical Engineering, ETH Zurich & University of Zurich, Zürich, Switzerland, 2Translational Neuromodeling Unit, University of Zurich & ETH Zurich, Zürich, Switzerland |
||
Standard fMRI acquires volumes as stacks of slices with 2D sequences. Acquisition time differences within a volume can be corrected post hoc using a sinc interpolation in image space (“slice timing correction”). In 3D sequences, timing correction is not possible in image space. Here, we verify an approach that rests on a sinc interpolation of timeseries in k-space before image reconstruction. Particularly, we investigate the artifacts which get removed by timing correction and quantify effects on the power spectra of timeseries in image space for 3D EPI. |
| 2342 | Computer 78
|
Model-based self-navigated water/fat decomposition for segmented diffusion-weighted EPI |
| Yiming Dong1, Malte Riedel2, Kirsten Koolstra3, Matthias J.P. van Osch1, and Peter Börnert1,4 | ||
1C.J. Gorter Center for High Field MRI, Department of Radiology, LUMC, Leiden, Netherlands, 2University and ETH Zurich, Zurich, Switzerland, 3Division of Image Processing, Department of Radiology, LUMC, Leiden, Netherlands, 4Philips Research, Hamburg, Germany |
||
Multi-shot EPI readout-approaches provide high spatial resolution at reduced geometric distortions and improved SNR in diffusion weighted imaging (DWI). As a specific challenge, multi-shot acquisition data require corrections for motion-induced, shot-specific phase errors, e.g. using additional navigator signals or appropriate self-navigation. Furthermore, proper fat-suppression is challenging in DWI, especially at B0 critical regions, making the use of chemical-shift encoding interesting. Therefore, an iterative, model-based reconstruction algorithm with self-navigation and water/fat decomposition, is proposed in this work. In-vivo examples in the leg and head-neck regions demonstrate improved water/fat separation as compared to acquisition-navigator approaches, while measurement times can be shortened. |
| 2343 | Computer 79
|
SENSE-based multipeak water/fat separation for diffusion-weighted imaging using self-navigated interleaved EPI |
| Yiming Dong1, Kirsten Koolstra2, Malte Riedel3, Matthias J.P. van Osch1, and Peter Börnert1,4 | ||
1C.J. Gorter Center for High Field MRI, Department of Radiology, LUMC, Leiden, Netherlands, 2Division of Image Processing, Department of Radiology, LUMC, Leiden, Netherlands, 3University and ETH Zurich, Zurich, Switzerland, 4Philips Research, Hamburg, Germany |
||
The presence of fat signals is one challenge for diffusion-weighted EPI, especially when considering the multi-peak spectrum nature of fat. In this work, we propose an improved SENSE-based water/fat separation algorithm to suppress multi-peak fat signals and apply this specifically to diffusion-weighted multi-shot EPI. The motion-induced shot-to-shot phase variations, an inevitable challenge in multi-shot DWI, are incorporated into the signal model using either a self-navigation or an extra-navigated method. The results show that the proposed SENSE-based algorithm yields good water/fat separation for non-diffusion and diffusion data with a multi-peak fat spectrum model. |
| 2344 | Computer 80
|
g-CAMP reconstructs multiple b-value diffusion weighted images from a single RESOLVE k-space |
| Nahla M H Elsaid1, Hemant D Tagare1,2, and Gigi Galiana1 | ||
1Radiology and Biomedical Imaging, Yale School of Medicine, New Haven, CT, United States, 2Department of Biomedical Engineering, Yale University, New Haven, CT, United States |
||
This work demonstrates the feasibility of reconstructing multiple b-value diffusion-weighted images (DWI) from a single RESOLVE k-space, where each blind was acquired with a different b-value. This multi b-value reconstruction uses the growing Constrained Alternating Minimization for Parameter mapping (g-CAMP) reconstruction method previously presented in ISMRM 2021. This can allow undersampling in the diffusion weighting dimension that is compatible with common undersampling schemes such as GRAPPA and multi-band EPI. |
| 2345 | Computer 81
|
Multi-domain Neumann Network with Sensitivity maps for Parallel MRI Reconstruction |
| Junhyeok Lee1, Junghwa Kang1, Se-hong Oh1, and Dong Hye Ye2 | ||
1Biomedical Engineering, Hankuk University of Foreign Studies, Yongin-si, Gyeonggi-do, Korea, Republic of, 2Department of Electrical and Computer Engineering, Marquette University, Milwaukee, WI, United States |
||
We performed parallel MRI reconstruction from under-sampled k-space data using the Multi-Domain Neumann Network with Sensitivity Maps. The Neumann network solves the inverse problem with recursive neural networks taking account into the forward model. We adapt the Neumann network with the coil sensitivity estimation and k-space data regurlaization to take account into MR physical models. Our proposed method shows uppressed image artifacts and enhanced spatial resolution compared with GRAPPA, U-Net and Neumann network. |
| 2346 | Computer 82
|
Model-based Image Reconstruction in Looping-star MRI |
| Haowei Xiang1, Jeffrey A Fessler1, and Douglas C Noll2 | ||
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, Ann Arbor, MI, United States, 2Biomedical Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, United States |
||
Looping star is a silent MRI pulse sequence that can be used for quantitative susceptibility mapping (QSM), T2*-weighted imaging and fMRI. The conventional reconstruction approach for looping star MRI that filters out some the k-space data does not fully model the overlapping echoes, and remove potentially useful signals. This work proposes a model-based reconstruction method that can theoretically resolve the overlapping echos and improve the SNR without increasing the scan time. |
| 2347 | Computer 83
|
Eliminating limits of spatiotemporal resolution in radial stack-of-stars imaging using FID navigators and single-readout binning |
| I. T. Maatman1, S. Ypma1, M. Kachelrieß2, Y. Berker2, K. T. Block3, E. Van der Bijl1, J. J. Hermans1, M. C. Maas1, and T. W. J. Scheenen1 | ||
1Radboud University Medical Centre, Nijmegen, Netherlands, 2German Cancer Research Center (DKFZ), Heidelberg, Germany, 3NYU Langone Medical Center, New York, NY, United States |
||
Motion-compensated images can be created from motion-binned undersampled radial stack-of-stars data through compressed sensing and image registration. However, for long repetition times or for many partitions, the acquisition time for one radial projection with all phase-encode steps becomes too long to sample the motion via self-gating, which leads to motion artifacts. Therefore, we estimate motion from FID-navigators and perform binning on a single-readout level to gain higher spatiotemporal resolutions. Our methods are tested on a motion phantom and volunteer with gridding and motion-compensated reconstructions. Our results show accurate detection of the motion signal and reduced motion blur in reconstructions. |
| 2348 | Computer 84
|
Low Latency Real-Time MRI at 0.55T using Self-Calibrating Through-Time GRAPPA |
| Prakash Kumar1, Yongwan Lim1, and Krishna S. Nayak1 | ||
1Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States |
||
Real-time MRI (RT-MRI) captures movements and dynamic processes in the human body as they occur, without reliance on any repetition or synchronization. Many applications of RT-MRI require low-latency reconstruction for feedback or interventions (typically <200ms). Here, we investigate self-calibrating spiral Through-Time GRAPPA at 0.55T and compare its performance against viewsharing and constrained reconstruction methods. We use longer readouts, which can be used at low-field due to reduced off-resonance effects. We demonstrate RT-MRI of speech production with 13.09ms temporal resolution and 25ms reconstruction latency (after a 30s pre-computed calibration step). |
| 2349 | Computer 85
|
Fast $$$T_{2}^{*}$$$ and QSM mapping using temporal CS with a Complementary Stochastic Sampling Scheme |
| Charles Iglehart1, Ali Bilgin1, and Manojkumar Saranathan2 | ||
1Electrical and Computer Engineering, University of Arizona, Tucson, AZ, United States, 2Department of Medical Imaging, University of Arizona, Tucson, AZ, United States |
||
We develop the motivation for and demonstrate the functionality of a temporal model-based complex multi-echo reconstruction algorithm coupled with a Complementary Stochastic Sampling procedure allowing for temporal sparsity to be leveraged . We demonstrate results for magnitude and phase images as well as parameter maps. |
| 2350 | Computer 86
|
Automatic reconstruction of arbitrary MRI sequences based on phase distribution graphs |
| Jonathan Endres1, Hoai Nam Dang2, Felix Glang3, Alexander Loktyushin3, Simon Weinmüller2, and Moritz Zaiss2,3 | ||
1Universitätsklinik Erlangen, Erlangen, Germany, 2Department of Neuroradiology, Universitätsklinik Erlangen, Erlangen, Germany, 3Magnetic Resonance Center, Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany |
||
We propose a method to automatically generate estimators for the exact encoding of a MRI signal, based on the Phase Distribution Graph of a sequence and its simulation. The estimator can then be used on a measurement to split the signal into its differently encoded parts, which enables reconstruction tailored to the sequence used. This approach can result in fewer imaging artifacts since it does not rely on any assumptions about the sequence made up front but on data obtained by a PDG simulation. This also makes it suitable for sequence optimization because it can adapt to changing sequence properties. |
| 2351 | Computer 87
|
Preclinical Evaluation of Two Compressed Sensing Methods for DTI |
| Diego Alves Rodrigues de Souza1, Hervé Mathieu1,2, Jean-Christophe Deloulme1, and Emmanuel L. Barbier1,2 | ||
1Univ. Grenoble Alpes - Inserm U1216 - Grenoble Institut des Neurosciences (GIN), Grenoble, France, 2Univ. Grenoble Alpes - Inserm US17 - CNRS UMS3552 - CHUGA - IRMaGe, Grenoble, France |
||
Compressed Sensing (CS) is not routinely implemented at the preclinical level, especially in case of multiple receivers. In this study, we evaluate preclinical Kernel Low-Rank and Conventional CS reconstruction schemes at 9.4T using a 4-channel receive coil and spin-echo data. By analyzing diffusion parametric maps and white-matter tracts, we evaluate the median absolute error, the structural similarity index measure and the mean fiber length as a function of the acceleration factor and of the CS reconstruction pipeline. |
| 2352 | Computer 88
|
Estimating B0 changes in Oscillating Steady State Imaging (OSSI) using an Artificial Neural Network |
| Mariama Salifu1, Melissa Haskell2, and Douglas C Noll1 | ||
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, 2Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States |
||
Oscillating steady state imaging (OSSI) is a novel fMRI acquisition method which produces a high SNR T2*-weighted signal. Physiological and drift induced B0 changes cause undesired signal distortions which can significantly diminish functional contrast in OSSI fMRI. Here, we investigated a method of rapidly estimating quantitative B0 field changes from OSSI images using an artificial neural network (ANN) model. Our results demonstrated that this technique can be used to rapidly measure field changes, which has the potential to be used for prospective and retrospective B0 field correction in OSSI. |
| 2353 | Computer 89
|
Realtime dynamic xyz-shimming in the cervical spinal cord |
| Daniel Papp1, Alexandre D’Astous1, Julien Cohen-Adad1,2,3, and Eva Alonso-Ortiz4 | ||
1NeuroPoly Lab, Institute of Biomedical Engineering, Polytechnique Montreal, Montreal, QC, Canada, 2Mila - Quebec AI Institute, Montreal, Montreal, QC, Canada, 3Functional Neuroimaging Unit, Centre de recherche de l'Institut universitaire de gériatrie de Montréal, Montreal, QC, Canada, 4NeuroPoly Lab, Department of Electrical Engineering, Polytechnique Montreal, Montreal, QC, Canada |
||
Respiration-induced changes of the magnetic field are an important source of image artefacts in spinal cord MRI. Here, we introduce realtime dynamic xyz-shimming to fully compensate for these distortions in a time-varying (‘realtime”), slicewise (‘dynamic”) fashion, using both in-plane (‘xy’) and through-plane (‘z’) gradients (‘xyz-shimming’). |
| 2354 | Computer 90
|
DeepFLAIR: a neural network approach to mitigate signal loss in temporal lobe regions of 7 Tesla FLAIR images |
| Daniel Uher1, Jacobus F.A. Jansen1,2, Gerhard S. Drenthen1,2, Benedikt A. Poser3, Christopher J. Wiggins4, Paul A.M. Hofman2,5, Louis G. Wagner5, Rick H.G.J. van Lanen6, Christianne M. Hoeberigs2,5, Albert J. Colon5, Olaf E.M.G. Schijns1,5,6, and Walter H. Backes1,2 | ||
1School for Mental Health and Neuroscience, Maastricht University, Maastricht, Netherlands, 2Department of Radiology & Nuclear Medicine, Maastricht University Medical Centre, Maastricht, Netherlands, 3Department of Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands, 4Scannexus, Maastricht, Netherlands, 5Academic Centre for Epileptology, Kempenhaeghe/Maastricht University Medical Centre, Heeze/Maastricht, Netherlands, 6Department of Neurosurgery, Maastricht University Medical Centre, Maastricht, Netherlands |
||
In this study we aimed to improve the 7T FLAIR image quality, especially within the temporal lobe regions which are often attenuated due to field inhomogeneities. A neural network using MP2RAGE and T2-weighted images as inputs was set up to generate a new FLAIR-like image. The training was performed on the extratemporal-lobe voxels of the acquired 7T FLAIR image. The deepFLAIR showed a significant improvement in the signal-to-noise ratio and contrast-to-noise ratio in the temporal lobe regions in a number of cases. This study showed the potential to generate FLAIR-like images with reduced inhomogeneity artifacts and improved image quality. |
| 2355 | Computer 91
|
Distortion correction in multi-echo MRI without field mapping or reverse encoding |
| Yael Balbastre1,2, Divya Varadarajan1,2, Klara Bas3, John Ashburner3, Robert Frost1,2, and Bruce Fischl1,2,4 | ||
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 2Department of Radiology, Harvard Medical School, Boston, MA, United States, 3Wellcome Center for Human NeuroImaging, Queen Square Institute of Neurology, University College London, London, United Kingdom, 4Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, United States |
||
Low-bandwidth multi-echo acquisitions suffer from spatial distortions between odd and even echoes caused by B0 homogeneity, causing blurs and loss of effective resolution in sum-of-square images or derived quantitative parameters. We show in this abstract that this curse can also be beneficial, as it allows distortions to be mapped without the need for additional field mapping data. We do so by jointly optimizing a forward model of the data with respect to exponential decay parameters and a smooth distortion map. |
| 2356 | Computer 92
|
Off-resonance correction in MRF using Deep Learning in fingerprint space |
| Ronal Coronado1,2, Carlos Castillo-Passi1,2, Gabriel della Maggiora1,2, Sergio Manuel Uribe1,2, Cristian Tejos1, Claudia Prieto1,2,3, Cecilia Besa4, and Pablo Irarrazaval1,2 | ||
1Biomedical Imaging Center-Universidad Católica de Chile, Santiago, Chile, 2Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, 3King's College London, London, United Kingdom, 4Departamento de Radiologia-Universidad Católica de Chile, Santiago, Chile |
||
Magnetic Resonance Fingerprinting (MRF) acquisitions with balanced Steady State Free Precession (bSSFP) and spiral trajectories are prone to off-resonance artifacts. These artifacts affect the reconstruction of the tissue maps (T1 and T2). We propose to use a UNet CNN feed with fingerprints corrupted by off-resonance to generate corrected fingerprints with only aliasing in the bSSFP-MRF sequence. The feasibility of the proposed approach was evaluated in simulations and in-vivo brain data. Our method improved the NRMSE values for both quantitative maps T1 and T2. Considerably reducing the effects of the off-resonance by UNet-MRF in comparison to classical bSSFP-MRF. |
| 2357 | Computer 93
|
Comparison of distortion correction with Echo Planar Imaging Correction and the 2D navigator based multi-shot SENSE EPI schemes |
| Yajing Zhang1, Zhigang Wu2, Xiuquan Hu3, Guangyu Jiang4, Jing Zhang3, Yan Zhao5, Marc Van Cauteren6, and Jiazheng Wang3 | ||
1BU-MR Clinical Science, Philips Healthcare, Suzhou, China, 2Philips Healthcare (China), Shenzhen, China, 3Philips Healthcare (China), Beijing, China, 4BU-MR Clinical Application, Philips Healthcare, Suzhou, China, 5BU-MR R&D, Philips Healthcare, Suzhou, China, 6BU-MR Clinical Science, Philips Healthcare, Tokyo, Japan |
||
Brain diffusion imaging by single shot EPI has remained challenging for the local B0 field variation, which leads to geometry distortion and signal inhomogeneity in the resulting images. In this study, we show that the brain single shot EPI image quality could be improved by correction of the geometric distortion and increased resolution, when compared to the traditional acquisition scheme. In-plane resolution of 1.2x1.2 mm2 was achieved and the distortion correction was effectively reduced. |
| 2358 | Computer 94
|
Measurements of composite signals with MR-based neuronal current imaging: one step closer to clinical application. |
| Milena Capiglioni1, Claus Kiefer2, and Roland Wiest2 | ||
1Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging, University of Bern, Bern, Switzerland, 2Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging (SCAN), Inselspital, Bern, Bern, Switzerland |
||
We evaluated the performance of stimulus-induced rotational saturation contrast (SIRS) to assess the possibility of detecting neuronal currents. We analyzed composite signals with multiple frequencies to evaluate the ability of the technique to be used as a filter of spectral components. We conclude that the method can separate and distinguish frequency components and reconstruct the distribution of spectral components between individually acquired signals. These tests are the basis for moving towards the clinical application of this sequence to locate fields with frequencies associated with specific pathologies. |
| 2359 | Computer 95
|
MEDI-FM: Field Map Error Guided Regularization for Shadow Reduction |
| Alexandra Grace Roberts1, Pascal Spincemaille2, Thanh Nguyen2, and Yi Wang2,3 | ||
1Electrical and Computer Engineering, Cornell University, Ithaca, NY, United States, 2Radiology, Weill Cornell Medicine, New York, NY, United States, 3Biomedical Engineering, Cornell University, Ithaca, NY, United States |
||
Morphology Enabled Dipole Inversion (MEDI) is an iterative reconstruction algorithm for Quantitative Susceptibility Mapping (QSM) that is effective in suppressing streaking artifacts by exploiting the magnitude image as a morphological prior. However, contiguous areas of dipole-incompatibility (such as noise) induce shadow artifacts whose spatial frequency components are not sufficiently regularized by the gradient based regularization in MEDI. Regularizing these spatially connected regions reduces shadow artifacts in Morphology Enabled Dipole Inversion based Quantitative Susceptibility Mapping reconstructions. |
| 2360 | Computer 96
|
B0 field distortions monitoring and correction for 3D non-Cartesian fMRI acquisitions using a field camera: Application to 3D-SPARKLING at 7T |
| Zaineb Amor1, Chaithya G.R.1,2, Caroline Le Ster1, Guillaume Daval-Frérot1,2,3, Nicolas Boulant1, Franck Mauconduit1, Christian Mirkes4, Philippe Ciuciu1,2, and Alexandre Vignaud1 | ||
1Université Paris-Saclay, CEA, Neurospin, CNRS, Gif-sur-Yvette, France, 2Université Paris-Saclay, Inria, Parietal, Palaiseau, France, 3Siemens Healthcare SAS, Saint Denis, France, 4Skope Magnetic Resonance Technologies AG, Zurich, Switzerland |
||
The benefit of retrospective correction of static and dynamic $$$B_{0}$$$ field imperfections was studied on non-Cartesian compressed sensing-based acquisition patterns by taking as an example 3D-SPARKLING encoding scheme. Long-TR dynamic acquisitions of T2*-weighted MRI (fMRI-like) volumes were acquired at 7T, retrospectively corrected, and the evaluation of the corrections was based on the image quality and the gain in temporal signal-to-noise ratio (tSNR). We found that correcting for the static and dynamic $$$0^{th}$$$ order contributions is enough to enhance the image quality and tSNR significantly (up to 22% gain in median tSNR). |
| 2433 | Computer 62
|
Accelerated Quantitative 3D UTE-Cones Imaging using Compressed Sensing |
| Jiyo S Athertya1, Ya-Jun Ma1, Amir Masoud Afsahi1, Alicia Ji1, Eric Y Chang1,2, Jiang Du1, and Hyungseok Jang1 | ||
1Radiology, University of California San Diego, San Diego, CA, United States, 2Radiology Service, VA San Diego Healthcare System, San Diego, CA, United States |
||
Quantitative ultrashort echo time (qUTE) imaging suffers from long acquisition time due to multiple acquisitions required for parameter estimation. In this study, feasibility of accelerated qUTE Cones imaging with compressed sensing (CS) reconstruction is investigated for fast variable flip angle UTE-T1 mapping, adiabatic UTE-T1ρ mapping, and UTE quantitative magnetization transfer (MT) modelling of macromolecular fraction (MMF). We explored these biomarkers for qUTE-Cones imaging of in vivo human knee joints at various undersampling rates. The performance of CS-reconstruction and parameter mapping was evaluated in tendons, ligaments, menisci, and cartilage. |
| 2434 | Computer 63
|
Multi-contrast Multi-scale vision Transformers (MMT) for MRI Contrast Synthesis |
| Jiang Liu1, Srivathsa Pasumarthi Venkata2, and Keshav Datta2 | ||
1Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, United States, 2R&D, Subtle Medical Inc, Menlo Park, CA, United States |
||
Complementary information from multi-modal MRI is widely used in clinical practice for disease diagnosis. Due to scan time limitations, image corruptions, and different acquisition protocols, one or more contrasts may be missing or unusable. Recently developed CNN models for contrast synthesis are unable to capture the intricate dependencies between input contrasts and are not dynamic to the varying number of inputs. This work proposes a novel Multi-contrast and Multi-scale vision Transformer (MMT) that can take any number and combination of input sequences and synthesize the missing contrasts. |
| 2435 | Computer 64
|
Open-source model-based reconstruction in Julia: A pipeline for spiral diffusion imaging |
| Alexander Jaffray1,2, Zhe (Tim) Wu1, Kamil Uludag3,4, and Lars Kasper1 | ||
1Techna Institute, University Health Network, Toronto, ON, Canada, 2UBC MRI Research Center, University of British Columbia, Vancouver, BC, Canada, 3Techna Institute & Koerner Scientist in MR Imaging, University Health Network, Toronto, ON, Canada, 4Biomedical Engineering, Center for Neuroscience Imaging Research, Institute for Basic Science, Sungkyunkwan University, Suwon, Korea, Republic of |
||
We investigated the use of physically-motivated corrections for non-Cartesian imaging within a fast, open-source MR reconstruction framework written in the Julia programming language. Building on existing Julia libraries for MR image reconstruction, we developed a comprehensive, open-source pipeline for non-Cartesian image reconstruction including B0 inhomogeneity correction, gradient system characterization and eddy current correction. We demonstrated that spiral images can be reconstructed using this pipeline with high geometric accuracy in a fast, accessible and efficient manner on modest hardware. |
| 2436 | Computer 65
|
Recurrent Variational Inference for fast and robust reconstruction of accelerated FLAIR MRI in Multiple Sclerosis |
| D. Karkalousos1, L. C. Liebrand1, S. Noteboom2, H. E. Hulst2,3, F. M. Vos4, and M. W. A. Caan1 | ||
1Department of Biomedical Engineering & Physics, Amsterdam UMC, University of Amsterdam, Amsterdam, Netherlands, 2Department of Anatomy & Neurosciences, MS Center Amsterdam, Amsterdam Neuroscience, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, Netherlands, 3Department of Medical, Health and Neuropsychology, Leiden University, Leiden, Netherlands, 4Delft University of Technology, Delft, Netherlands |
||
Robustness when applying Deep Learning methods to clinical data is crucial for accurate high-resolution reconstructions while having fast inference times. We propose the Cascades of Independently Recurrent Variational Inference Machine (CIRVIM), targeting deep unrolled optimization and enforcing data consistency for further robustness. We quantify contrast resolution of seven and half times prospectively undersampled FLAIR MRI without fully-sampled center containing Multiple Sclerosis lesions. The proposed scheme reduces inference times by a factor of 6 compared to Compressed Sensing. Lesion contrast resolution improves by approximately 13% while preserving spatial detail with enhanced sharpness compared to more blurred results of other presented methods. |
| 2437 | Computer 66
|
Training a tunable, spatially-adaptive denoiser without clean targets |
| Laura Pfaff1,2, Julian Hossbach1,2, Elisabeth Preuhs1, Tobias Wuerfl2, Silvia Arroyo Camejo2, Dominik Nickel2, and Andreas Maier1 | ||
1Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany, 2MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany |
||
Accelerating MRI is intrinsically limited by the thermal noise from the imaged object. In this work we aim to optimize MR image denoising using an unsupervised deep learning-based method. Stein's unbiased risk estimator and spatially resolved noise maps indicating the standard deviation of the noise for every pixel were incorporated into the training process. It was shown that this approach can achieve results that are equal or superior to those of state-of-the-art supervised and unsupervised methods. Furthermore, we show how to control the tradeoff between denoising and image sharpness by using a model conditioned on the noise map. |
| 2438 | Computer 67
|
Explainable multi-contrast deep learning model with anomaly-aware attention for reduced gadolinium dose in CE brain MRI - a feasibility study |
| Srivathsa Pasumarthi Venkata1, Ben Andrew Duffy1, Enhao Gong2, Greg Zaharchuk3, and Keshav Datta1 | ||
1R&D, Subtle Medical Inc, Menlo Park, CA, United States, 2R&D, Subtle Medical Inc., Menlo Park, CA, United States, 3Department of Radiology, Stanford University, Stanford, CA, United States |
||
Complementary information from multi-contrast MRI data is used in deep learning algorithms for reducing contrast dosage in brain MRI. Though existing models produce clinically equivalent post-contrast images, they lack explainability in terms of mapping the source of contrast information from input to output. In this work we explore the feasibility of an explainable deep learning model for gadolinium dose reduction in contrast-enhanced brain MRI. |
| 2439 | Computer 68
|
A Least-squares Method for Improved Sensitivity Map Estimation when Imaging X-Nuclei |
| Nicholas Dwork1, Shuyu Tang1, Peder E. Z. Larson1, and Jeremy W. Gordon1 | ||
1Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States |
||
We present a least-squares method for estimating sensitivity when imaging x-nuclei. We show results with hyperpolarized pyruvate in both the heart and the brain. |
| 2440 | Computer 69
|
Rapid high-resolution cranial bone MRI using deep-learning prior image reconstruction |
| Parna Eshraghi Boroojeni1, Paul Commean1, Cihat Eldeniz1, Weijie Gan1, Gary Skolnick1, Kamlesh Patel1, Ulugbek Kamilov1, and Hongyu An1 | ||
1Washington University in Saint Louis, Saint Louis, MO, United States |
||
A high-resolution (HR) MRI capable of resolving the detail of bony structures at sub-millimeter resolution is desired. A short MR acquisition results in under-sampled k-space data below the Nyquist rate, leading to artifacts and high noise. We developed an HR reconstruction method regularized by a complex deep-learning prior (RECD). We achieved high-resolution MR (0.6x0.6x0.8mm3) with a one-minute acquisition time. Using images reconstructed from a 5-minute MR scan as the gold standard, we compared the peak signal to noise ratio (PSNR) and similarity index (SSIM) for 1-min RECD and 1-min compressed sensing (CS) reconstructed images. RECD outperformed CS. |
| 2441 | Computer 70
|
Comparison between GRAPPA and automatic compressed sensing-based reconstruction up to 16-fold acceleration |
| Gabriel Varela-Mattatall1,2, Roy A.M. Haast 1,3, Omer Oran4, Ali R. Khan 1,2, and Ravi S. Menon1,2 | ||
1Centre for Functional and Metabolic Mapping (CFMM) | Robarts Research Institute | Western University, London, ON, Canada, 2Department of Medical Biophysics | Schulich School of Medicine and Dentistry | Western University, London, ON, Canada, 3Aix-Marseille Universite | CNRS | CRMBM, Marseille, France, 4Siemens Healthcare Limited, Oakville, ON, Canada |
||
With the recent development of an automatic compressed sensing (CS)-based reconstruction method, we investigated its allowance to recover T2* maps from prospectively undersampled multi-echo, gradient echo-based acquisitions at 7T. We compared our CS method with the more conventional GRAPPA-based reconstruction using up to 16-fold acceleration. In contrast to GRAPPA, our CS-based method allows recovery of T2* maps up to 16-fold acceleration, which demonstrates its promise for ultra-fast acquisitions. However, current results also show some caveats that need to be addressed as future work. |
| 2442 | Computer 71
|
Coil Selective Golden Angle DCE-MR Image Reconstruction using Mutual Dependence |
| Aziz Kocanaogullari1, Cemre Ariyurek1, Onur Afacan1, and Sila Kurugol1 | ||
1Radiology, Boston Children's Hospital, Boston, MA, United States |
||
MRI literature shows that it is possible to increase image reconstruction quality by removing coils that cause majority of the streaking artifacts. We introduce to measure the coil quality based on mutual information between a reference and each coil image. Specifically we apply this technique to DCE-MRI reconstruction from under-sampled radial stack of stars trajectory for kidney imaging and calculate mutual dependence between a dynamic image from each coil and a reference image to assess coil contribution to reconstruction. Experiments show mutual dependence based coil selection reduces artifacts and increases reconstructed image SNR by $$$16.84\%$$$ using $$$1/3$$$ of the coils. |
| 2443 | Computer 72
|
Reconstruction for Simultaneous Multi-Slab Spatiotemporal Encoding (SMS SPEN) Using Split Slice-GRAPPA combined with CAIPIRINHA |
| Jaeyong Yu1,2, Sugil Kim3, and Jang-Yeon Park1,2 | ||
1Department of Biomedical Engineering, Sungkyunkwan University, Suwon, Korea, Republic of, 2Department of Intelligent Precision Healthcare Convergence, Sungkyunkwan University, Suwon, Korea, Republic of, 3Siemens Healthineers Ltd., Seoul, Korea, Republic of |
||
To reduce acquisition time, acceleration techniques such as simultaneous multi-slice and parallel imaging techniques have been applied to various sequences. In this work, we proposed a reconstruction method for simultaneous multi-slab spatiotemporal-encoding (SMS SPEN) using split slice-GRAPPA combined with CAIPIRINHA. The proposed method was applied to the recently developed ultrafast 3D SPEN imaging sequence called ERASE (Equal-TE Rapid Acquisition with Sequential Excitation) and a case of multiband factor of 3 was demonstrated with human brain imaging at 3T. |
| 2444 | Computer 73
|
Abdominal 3D-T1-weighted imaging with the radial MP2RAGE sequence. |
| François Emmanuel Jean Guy MAINGAULT1, Nadège CORBIN1, William LEFRANÇOIS1, Aurélien J TROTIER1, Sylvain MIRAUX1, and Emeline J RIBOT1 | ||
1Centre de Résonance Magnétique des Systèmes Biologiques, UMR5536, CNRS, Bordeaux, France, Metropolitan |
||
The Magnetization Prepared Two RApid Gradient Echo (MP2RAGE) sequence with a cartesian sampling is performed to obtain 3D-T1 weighted images of the brain. A radial encoding giving a better robustness to respiratory motion has been implemented in order to use the MP2RAGE sequence for abdominal imaging. |
| 2445 | Computer 74
|
Calibration kernels with Alternative Sampling Scheme (CASS) for Parallel Imaging: SENSE meets CASS |
| Sangcheon Choi1,2, Rahel Heule1,3, Edyta Leks1,2,3, Felix Glang1, Xin Yu4, and Klaus Scheffler1,3 | ||
1Department of High-field Magnetic Resonance, Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, 2Graduate Training Centre of Neuroscience, University of Tuebingen, Tuebingen, Germany, 3Department of Biomedical Magnetic Resonance, University of Tuebingen, Tuebingen, Germany, 4Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Harvard Medical School, Massachusetts General Hospital, Charlestown, MA, United States |
||
We developed new calibration kernels with an alternative undersampling scheme (CASS) for parallel imaging to reduce coherent aliasing artifacts and noises. By sampling k-space lines with irregular and blockwise patterns, incoherent aliasing patterns and noise signals were spread in reconstructed CASS images. Noteworthily, the CASS method outperformed the conventional GRAPPA method at higher acceleration factors. |
| 2446 | Computer 75
|
BOLD PSF: Impact of k-space sampling on T2* contrast |
| Maria Engel1, Lars Kasper1,2, and Klaas Paul Pruessmann1 | ||
1Institute for Biomedical Engineering, ETH Zurich and University of Zurich, Zurich, Switzerland, 2Translational Neuromodeling Unit, Institute for Biomedical Engineering, University of Zurich and ETH Zurich, Zurich, Switzerland |
||
The timing of BOLD-fMRI is typically chosen such that TE matches T2*. This assumes contrast being exclusively elicited by signal differences at TE. While sensible for short or TE-symmetric readouts, such as EPIs, it oversimplifies contrast generation for longer or TE-asymmetric trajectories, such as spirals. We propose the concept of a BOLD PSF for a more comprehensive perspective on the imaging characteristics of a functional experiment. Our findings indicate that TE can be reduced for spiral-out without sacrificing BOLD-sensitivity when compared to EPI. Furthermore, we characterize the intrinsic trade-off between specificity and resolution of the BOLD response under varying TE. |
| 2447 | Computer 76
|
Synchronizing TMS pulses for free artifact fMRI for multichannel TMS systems using the integrated whole head TMSMR 28 channel RF coil array |
| Lucia Navarro de Lara1,2, Isil Uluc1,3, Qinglei Meng1,3, Jason Stockmann1,3, Larry Wald1,3, and Aapo Nummenmaa1,3 | ||
1Martinos Center - MGH, Charlestown, MA, United States, 2Harvard Medical School, Boton, MA, United States, 3Harvard Medical School, Boston, MA, United States |
||
For TMS/fMRI experiments, to acquire functional images without artifacts, general recommendations have been proposed. In order to apply TMS pulses in a multichannel Transcranial Magnetic Stimulation system integrated with the new constructed TMSMR 28- channel RF coil, we have explored the different options and artifacts that can be generated depending on when the TMS pulses are applied with respect to the sequence timing. As conclusion, we propose a method using an RF pick up loop to assure that the delivery of the TMS pulses are not close to the EPI navigators or the imaging gradients. |
| 2448 | Computer 77
|
Magnetization transfer optimization to enhance 3D EPI image contrast for high-resolution fMRI applications at 7T |
| Vahid Malekian1, Nadine N. Graedel1, Oliver Josephs1, and Martina F. Callaghan1 | ||
1Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom |
||
3DEPI is widely used for cortical depth-dependent fMRI studies at 7T. However, it suffers from poor image contrast that can be problematic for fMRI post-processing steps. One solution to bolster contrast is to implement a magnetization transfer (MT) module. Here, we experimentally investigated the different MT pre-pulse characteristics to optimize image contrast, while considering power limitations at 7T. Contrast gains eased co-registration to an MP2RAGE reference, and improved cortical segmentation. Our analyses confirmed that the MT contrast is spatially variable and dependent on transmit field inhomogeneity.
|
| 2449 | Computer 78
|
Millimeter Spatial Resolution and Subsecond Temporal Resolution in Real-Time fMRI using Multi-Band Echo-Volumar Imaging |
| Stefan Posse1, Sudhir Ramanna2, Steen Moeller2, Bruno Sa De La Rocque Guimaraes1, Michael Mullen2, and Essa Yacoub2 | ||
1Neurology, U New Mexico, Albuquerque, NM, United States, 2Center for Magnetic Resonance Research, Minneapolis, MN, United States |
||
In this study we develop highly undersampled multi-band slab-segmented echo-volumar imaging (MB_EVI), which combines the sampling efficiency of single-shot 3D encoding with the sensitivity advantage of multi-echo acquisition, and explore the feasibility of shortening the readout duration using segmented within-slab slice encoding. Kz-segmented Segmented MB-EVI enabled a nominal 1x1x1 mm isotropic voxel size with 64 slices and a temporal resolution of 618 ms with sub-second acquisition time. |
| 2450 | Computer 79
|
Optimal Parameter Selection for 3D Radial Density Adapted Trajectories |
| Mark Bydder1, Fadil Ali1, Andres Saucedo1, Chencai Wang1, Akifumi Hagiwara1, Alex D Pham1, Jingwen Yao2, and Ben Ellingson1 | ||
1UCLA, Los Angeles, CA, United States, 2UCSF, San Francisco, CA, United States |
||
The present article develops an understanding of how to choose optimal (or at least reasonable) parameter values for 3D radial density adapted sampling. This has a practical role in designing clinical protocols for short TE non-Cartesian imaging. |
| 2451 | Computer 80
|
Prospects of non-Cartesian 3D-SPARKLING encoding for functional MRI: A preliminary case study for retinotopic mapping |
| Zaineb Amor1, Chaithya G.R.1,2, Guillaume Daval-Frérot1,2,3, Bertrand Thirion1,2, Franck Mauconduit1, Philippe Ciuciu1,2, and Alexandre Vignaud1 | ||
1Université Paris-Saclay, CEA, Neurospin, CNRS, Gif-sur-Yvette, France, 2Université Paris-Saclay, Inria, Parietal, Palaiseau, France, 3Siemens Healthcare SAS, Saint-Denis, France |
||
For the first time, non-Cartesian 3D-SPARKLING (Spreading Projection Algorithm for Rapid K-space sampLING) encoding scheme was used for task-based fMRI (retinotopic mapping) at 7T. Additionally, this new acquisition technique was compared with 3D-EPI, considered as one of the reference acquisition schemes in high-resolution fMRI. The experiment was performed on a single participant at 1mm isotropic and for a temporal resolution of 2.4s. We found that both techniques yield similar statistically significant activation patterns in the visual cortex. |
| 2452 | Computer 81
|
Channel Pruning to Improve Robustness of Pilot Tone-based Cardiac Trigger Detection |
| Jacob Hatef1, Abhishek Vijaykumar1, Chong Chen1, Yingmin Liu1, and Rizwan Ahmad1 | ||
1The Ohio State University, Columbus, OH, United States |
||
In this work, we develop and apply a method to prune channels from multi-coil Pilot Tone (PT) data. The method suppresses contributions from noisy or corrupt channels and thus improves the quality of the PT-driven cardiac signal. The method, called Spectrum Optimized Channel pruning (SOC), is based on maximizing the ratio of the signal energy in the frequencies that correspond to physiological motions to the signal energy in the frequencies that are outside the physiological motions. Using data from 15 subjects, we highlight that the proposed channel pruning method improves the quality of the extracted cardiac signal. |
| 2453 | Computer 82
|
Adaptive k-Space Sampling in Magnetic Resonance Imaging Using Reinforcement Learning |
| Edith Franziska Baader1, Fabian Theißen1, Nicola Pezzotti2,3, and Volkmar Schulz1,4,5,6 | ||
1Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany, 2Philips Research, Eindhoven, Netherlands, 3Department of Mathematics and Computer Science, Eindhoven University of Technology, Eindhoven, Netherlands, 4Physics Institute III B, RWTH Aachen University, Aachen, Germany, 5Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany, 6Hyperion Hybrid Imaging Systems GmbH, Aachen, Germany |
||
One approach to accelerate MRI scans is to acquire fewer k-space samples. Commonly, the sampling pattern is selected before the scan, ignoring the sequential nature of the sampling process. A field of machine learning addressing sequential decision processes is reinforcement learning (RL). We present an approach for creating adaptive two-dimensional (2D) k-space trajectories using RL and the so-called action space shaping. The trained RL algorithm adapts to a variety of basic 2D shapes outperforming simple baseline trajectories. By shaping the action space of the RL agent we achieve better generalization and interpretability of the agent. |
| 2454 | Computer 83
|
Compression sensing based uterine 3D-T2WI isotropic imaging: Optimization of the optimal accelerator |
| Changjun Ma1, Shifeng Tian2, Lihua Chen2, Nan Wang2, Qingwei Song2, Liangjie Lin3, and Ailian Liu2 | ||
1Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian,China, China, 2Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian, China, 3Philips Healthcare, Beijing, China, Beijing, China |
||
3D-ISO-T2WI can obtain ISO volume images compared with traditional 2D-T2WI imaging, reducing partial volume effects, and volume data post-processing can be omni-directional, Three-dimensional, multi-angle display of the overall picture of the uterine structure and the scope of the lesion, but its scanning time is longer [1], the peristalsis of the subject's uterus, bladder and intestine during the scanning time may affect the image quality, and the scanning time The long-term reduction of the patient’s coordination and examination circulation.In this study, different AF will be selected to optimize the best AF based on the CS-based 3D-ISO-T2WI sequence of the uterus. |
| 2455 | Computer 84
|
Single shot multiple overlapping-echo acquisition combined with deep learning for ultra-fast T2 mapping in glioma: A preliminary study |
| Jianfeng Bao1, Xiaoyue Ma2, Qinqin Yang3, Xiao Wang2, Yong Zhang2, Liangjie Lin4, Jingliang Cheng2, and Congbo Cai3 | ||
1The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China, 2Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China, 3Department of Electronic Science, Xiamen University, Xiamen, China, 4Philips Healthcare, Beijing, China |
||
Quantitative MRI (qMRI) are expected largely to provide more accuracy clinical diagnostic information. However, bound by the relative long duration for the routine MRI exam and much longer qMRI acquisition time, the qMRI is not widely used for diagnostic purpose. The change of MR signal caused by contrast agent extravasation is affected by many factors, which will increase difficulty in diagnosis. Quantitative MRI (qMRI) are expected to overcome this issue. Recently, an ultra-fast multi-parameters qMRI technique was developed by our group, and herein we aim to apply the proposed strategy on brain tumor imaging and to preliminary access the potential performance. |
| 2507 | Computer 39
|
Reducing streak artefacts in radial MR fingerprinting of the prostate through automated channel removal |
| Kaia Ingerdatter Sørland1, Cristopher George Trimble1, Elise Sandsmark2, Tone Frost Bathen1,2, Mattijs Elschot1,2, and Martijn A. Cloos3 | ||
1Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway, 2Department of Radiology and Nuclear Medicine, St. Olavs hospital, Trondheim University Hospital, Trondheim, Norway, 3Centre for Advanced Imaging, The University of Queensland, Brisbane, Australia |
||
The extreme undersampling factors used in radial magnetic resonance fingerprinting (MRF) of the prostate lead to strong streak artefacts from the femoral arteries and veins. However, it turns out that only a subset of receiver channels is responsible for these streaks. In this work we automatically detected and removed these from the reconstruction pipeline. The method was applied to MRF acquired with various acceleration factors in seven asymptomatic volunteers, significantly reducing visible streaks in reference tissues surrounding the prostate without impairing the prostate T1 and T2 values in the MRF maps. |
| 2508 | Computer 40
|
Quantification of microvascular properties of gliomas using DSC – Hybrid EPI based MR Vascular Fingerprinting compared with Vessel Size Imaging |
| Krishnapriya Venugopal1, Fatemehsadat Arzanforoosh1, Daniëlle van Dorth2, Marion Smits1, Juan Antonio Antonio Hernandez-Tamames1,3, Esther A.H Warnert1, Matthias J.P van Osch2, and Dirk H.J Poot1 | ||
1Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands, 2Radiology, Leiden University Medical Center, Leiden, Netherlands, 3Medical Imaging, TU Delft, Delft, Netherlands |
||
This study uses an MRVF approach to analyze the time evolution of a DSC hybrid -EPI (HEPI) sequence that simultaneously acquires gradient and spin echo images. HEPI properties are incorporated from the scanner into simulations including contrast agent extravasation, diffusion, and MR signal evolution and for varying the outcome parameters: CBV, mean vessel-size, and leakage. In vivo data of six glioma patients are used to compare MRVF output maps to those obtained from conventional VSI modelling. The results show reasonable agreement. Also, the noise sensitivity of both techniques was investigated. |
| 2509 | Computer 41
|
Spectral principal axes system (SPAS) and tuning of tensor-valued encoding for time-dependent anisotropic diffusion |
| Samo Lasič1,2, Nathalie Just1, and Henrik Lundell1 | ||
1Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Amager and Hvidovre, Danish Research Centre for Magnetic Resonance, Copenhagen, Denmark, 2Random Walk Imaging, Lund, Sweden |
||
Tensor-valued diffusion encoding can be confounded by time-dependent diffusion (TDD). Matching sensitivity to TDD or tuning of b-tensors with different shapes is needed for unbiased microstructure assessment. We present a method for tuning linear tensor encoding (LTE) to spherical tensor encoding (STE), which could be optimized for different hardware constraints. Furthermore, we introduce the spectral principle axes system (SPAS), representing spectral anisotropy of STE. The SPAS LTEs could provide an alternative to tuning and enable disentangling effects of microscopic anisotropy and TDD, useful to correlate cell shape and size. |
| 2510 | Computer 42
|
Tailored MR fingerprinting of pediatric brain tumor patients |
| Pavan Poojar1,2, Enlin Qian1, Alexis B Maddocks3, and Sairam Geethanath1 | ||
1Columbia Magnetic Resonance Research Center, Columbia University, New York, NY, United States, 2Dayananda Sagar College of Engineering, Bangalore, India, 3Columbia University Irving Medical Center, New York, NY, United States |
||
We demonstrated the utility of tailored MR fingerprinting (TMRF) of pediatric patients with brain tumor to 1) differentiate tumor from healthy tissue using quantitative maps; 2) tailor TMRF sequence to include T2 fluid-attenuated inversion recovery (FLAIR) contrast that could potentially substitute T2 FLAIR sequence used in routine pediatric tumor protocol. The mean土SD difference in T1 and T2 values between tumor and normal tissues for 3 patients were 1.05土0.09 and 0.05土0.02 seconds respectively. The TR and flip angle trains were tailored to include T2 FLAIR contrast for adult imaging. Current and future work includes optimizing T2 FLAIR for the pediatric population. |
| 2511 | Computer 43
|
Feasibility of Quantifying Cerebral Blood Volume and Blood-Brain Barrier Water Exchange using Non-Contrast MR Fingerprinting |
| Emma L Thomson1,2, Elizabeth Powell3, Claudia A M Gandini Wheeler-Kingshott2,4,5, and Geoff J M Parker1,2,6 | ||
1Centre for Medical Image Computing, Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 2NMR Research Unit, Queen Square MS Centre, Department of Neuroinflammation, UCL Queen Square Institute of Neurology, Faculty of Brain Sciences, London, United Kingdom, 3Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, 4Department of Brain & Behavioural Sciences, University of Pavia, Pavia, Italy, 5Brain Connectivity Centre Research Department, IRCCS Mondino Foundation, Pavia, Italy, 6Bioxydyn Limited, Manchester, United Kingdom |
||
We propose the use of magnetic resonance fingerprinting (MRF), applied using a spoiled gradient echo sequence, to quantify cerebral blood volume ($$$\nu_b$$$) and inter-vascular water exchange (1/$$$\tau_b$$$), without the need for contrast agents. Through a simulation study we optimise a simulated acquisition protocol and test the sensitivity of the measurement, and its accuracy in the presence of variations in blood T1, tissue T1, and B1. We demonstrate that voxel-wise simultaneous quantification of $$$\nu_b ,\tau_b , T_{1,b}, T_{1,e}$$$ and $$$B_1^+$$$ is likely feasible with an optimised acquisition. |
| 2512 | Computer 44
|
Evaluation of multiband variable-rate selective excitation (MB-VERSE) diffusion-weighted imaging (DWI) of the liver MRI |
| Ja Kyung Yoon1, Yong Eun Chung1, Jaeseung Shin1, Jin-Young Choi1, Mi-Suk Park1, and Myeong-Jin Kim1 | ||
1Radiology, Severance Hospital, Research Institute of Radiological Science, Yonsei University College of Medicine, Seoul, Korea, Republic of |
||
In liver MRI, diffusion-weighted imaging (DWI) allows better detection and characterization of focal lesions, but requires a relatively long scan time. The multiband variable-rate selective excitation (MB-VERSE) echoplanar imaging for DWI provides accelerated acquisition time with some expected trade-off in image quality. Qualitative and quantitative image quality of MB-VERSE images as well as focal liver lesion detectability was evaluated by three readers. The MB-VERSE sequence as well as focal liver lesion detectability showed significant sacrifice in quantitative and qualitative overall image quality, but comparable focal lesion detectability. |
| 2513 | Computer 45
|
Combining Region-optimized virtual coils with low rank reconstructions for accelerated cardiac MR Fingerprinting |
| Gastao Cruz1, Andrew Phair1, Carlos Velasco1, René M. Botnar1, and Claudia Prieto1 | ||
1King's College London, London, United Kingdom |
||
Cardiac Magnetic Resonance Fingerprinting (MRF) produces co-registered, multi-parametric maps from highly accelerated acquisitions. Low rank methods leveraging temporal information have enabled highly undersampled MRF reconstructions. However, residual aliasing from these reconstructions can potentially propagate through the MRF framework into the parametric maps. Region-Optimized Virtual coils have recently been proposed for undersampled reconstructions in conventional imaging, by suppressing unwanted sources of aliasing signal. Here we combine both methods and investigate its performance in Cardiac Magnetic Resonance Fingerprinting. |
| 2514
|
Computer 46
|
Joint sparsity multi-component MRF reconstruction - directly from k-space to component maps |
| Martijn Nagtegaal1, Emiel Hartsema1, Kirsten Koolstra2, and Frans Vos1,3 | ||
1Imaging Physics, Delft University of Technology, Delft, Netherlands, 2Department of Radiology, Leiden University Medical Centre, Leiden, Netherlands, 3Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands |
||
The use of high undersampling factors and short flip angle trains leads to shorter acquisition times in MR Fingerprinting acquisitions. To obtain accurate multi-component estimates from this data advanced reconstructions are required. We study a low-rank ADMM based reconstruction method that adds a multi-component constraint to the inverse reconstruction problem (MC-ADMM). This method is combined with a joint-sparsity constraint yielding higher quality multi-component estimates with k-SPIJN than with previous methods. In simulations we observed increased stability to sequence truncation and in vivo multi-component estimates contained less noise-like effects. |
| 2515 | Computer 47
|
Magnetic Resonance Fingerprinting with Total Nuclear Variation Regularisation |
| Imraj Ravi Devia Singh1, Olivier Jaubert1, Bangti Jin1, Kris Thielemans2, and Simon Arridge1 | ||
1Department of Computer Science, University College London, London, United Kingdom, 2Institute of Nuclear Medicine, University College London, London, United Kingdom |
||
Magnetic Resonance Fingerprinting (MRF) accelerates quantitative magnetic resonance imaging. The reconstruction can be separated into two problems: reconstruction of a set of multi-contrast images from k-space signals, and estimation of parametric maps from the set of multi-contrast images. In this study we focus on the former problem, while leveraging dictionary matching for the estimation of parametric maps. Two different sparsity promoting regularisation strategies were investigated: contrast-wise Total Variation (TV) which encourages image sparsity separately; and Total Nuclear Variation (TNV) which promotes a measure of joint edge sparsity. We found improved results using joint sparsity. |
| 2516 | Computer 48
|
3D magnetic resonance fingerprinting at 50 mT with integrated estimation and correction of image distortions due to B0 inhomogeneities |
| Thomas O'Reilly1, Peter Börnert2, Andrew Webb1, and Kirsten Koolstra1 | ||
1Leiden University Medical Center, Leiden, Netherlands, 2Philips Research Hamburg, Hamburg, Germany |
||
Strong B0 inhomogeneities in compact point-of-care low field permanent magnet systems can cause image distortions and reduced signal for large field-of-view images. MRF, as a flexible acquisition framework, can encode these effects into the acquisition process, to incorporate them into the reconstruction process. In this work, we use an alternating TE in the MRF sequence that supports ΔB0 estimation and image distortion correction of the MRF data. This method reduces distortions in the relaxation time maps without needing to acquire an additional ΔB0 map. |
| 2517 | Computer 49
|
Learning-based prediction of encoding capability for acquisition schedule of Magnetization Transfer Contrast MR fingerprinting |
| Beomgu Kang1, Hye-Young Heo2,3, and Hyunwook Park1 | ||
1School of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, 2Division of MR Research, Department of Radiology, Johns Hopkins University, Baltimore, MD, United States, 3F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States |
||
Magnetization transfer contrast MR fingerprinting (MTC-MRF) is used to quantify multiple tissue parameters of free bulk water and semisolid macromolecule using pseudo-randomized MRF schedules. An optimal design of the MRF schedule is important to improve the quantification accuracy and reduce the scan time. However, lack of the objective function that represents the encoding capability of MRF schedule hinders the reliable optimization. In this study, we propose a novel metric that represents the encoding capability of MRF schedule based on recurrent neural network. Unlike the conventional metrics based on indirect measurements, the proposed learning-based metric directly measures the tissue quantification errors. |
| 2518 | Computer 50
|
Application of Deep Learning techniques to Magnetic Resonance Fingerprinting |
| Raffaella Fiamma Cabini1,2, Leonardo Barzaghi1,3, Davide Cicolari2,4, Anna Pichiecchio5,6, Silvia Figini2,7, Paolo Arosio8,9, Marta Filibian2,10, Alessandro Lascialfari2,4, and Stefano Carrazza8,9 | ||
1Department of Mathematics, University of Pavia, Pavia, Italy, 2INFN, Istituto Nazionale di Fisica Nucleare - Pavia Unit, Pavia, Italy, 3Department of Neuroradiology, Advanced Imaging and Radiomics, IRCCS Mondino Foundation, Pavia, Italy, 4Department of Physics, University of Pavia, Pavia, Italy, 5Department of Neuroradiology, IRCCS Mondino Foundation, Pavia, Italy, 6Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, 7Department of Social and Political Science, University of Pavia, Pavia, Italy, 8Department of Physics, University of Milano, Milano, Italy, 9INFN, Istituto Nazionale di Fisica Nucleare - Milano Unit, Milano, Italy, 10Centro Grandi Strumenti, University of Pavia, Pavia, Italy |
||
We developed a Neural Network (NN) for the reconstruction of T1 and T2 parametric maps obtained with the Magnetic Resonance Fingerprinting (MRF) technique. The training phase was realized on experimental inputs, eliminating the use of simulated datasets and theoretical models. The set of optimal hyperparameters of the NN and the supervised training algorithm were established through an optimization procedure. The model achieved similar performances to the traditional reconstruction method, but the number of MRF images required was lower with respect to the dictionary-based method. If translated to the clinic, our results envisage a significant time shortening of MRI investigation. |
| 2519 | Computer 51
|
Through-plane Motion in Magnetic Resonance Fingerprinting: Simulations and Phantom Experiments |
| Martijn Nagtegaal1, Charles McGrath2, Christian Günthner2, and Manuel Bauman3 | ||
1Imaging Physics, Delft University of Technology, Delft, Netherlands, 2Institute for Biomedical Enginering, ETH Zurich, Zurich, Switzerland, 3Philips Research Europe, Hamburg, Germany |
||
- |
| 2520 | Computer 52
|
A Scalable High-Performance Magnetic Resonance Fingerprinting Search Engine using GPUs |
| Gabriel Zihlmann1, Najat Salameh1, and Mathieu Sarracanie1 | ||
1Department of Biomedical Engineering, University of Basel, Allschwil, Switzerland |
||
Dictionary search for the reconstruction of MR Fingerprinting-based sequences can become a bottleneck due to the exponential growth of dictionary size with the number of parameters. Several approaches have been proposed to accelerate this search, yet the availability of efficient and scalable software implementations remains limited. Here, we extended an existing, highly optimized similarity search library to be compatible with the requirements of MR Fingerprinting dictionary search. The evaluation of our GPU implementation shows an acceleration by two orders of magnitude over the brute-force search yielding identical results for 99.9 % of the voxels. |
| 2521 | Computer 53
|
Predicting PDFF and R2* from Magnitude-Only Two-Point Dixon MRI Using Generative Adversarial Networks |
| Nicolas Basty1, Marjola Thanaj1, Madeleine Cule2, Elena P. Sorokin2, E. Louise Thomas1, Jimmy D. Bell1, and Brandon Whitcher1 | ||
1Research Centre for Optimal Health, University of Westminster, London, United Kingdom, 2Calico Life Sciences LLC, South San Francisco, CA, United States |
||
We implement generative adversarial network (GAN) models to predict fully quantitative parameters from a complex-valued multiecho MRI sequence using only data from the magnitude-only two-point Dixon acquisition in the UK Biobank abdominal protocol. The training data consists of in- and opposed-phase channels from the Dixon sequence as inputs and the proton density fat fraction (PDFF) and R2* parameter maps estimated from the IDEAL acquisition as outputs. We compare conditional and cycle GANs, where the conditional GAN models outperformed the cycleGAN models in SSIM, PSNR and MSE. PDFF predictions were better than R2* predictions for all models. |
| 2522 | Computer 54
|
Fast and High-Resolution T2* weighted and susceptibility-based MRI using 3D-EPI with Deep Learning Reconstruction |
| Brice Fernandez1, R. Marc Lebel2, Xinzeng Wang3, Arnaud Guidon4, Suchandrima Banerjee5, Stefan Skare6,7, and Tim Sprenger8 | ||
1GE Healthcare, Buc, France, 2GE Healthcare, Calgary, AB, Canada, 3GE Healthcare, Houston, TX, United States, 4GE Healthcare, Boston, MA, United States, 5GE Healthcare, Menlo Park, CA, United States, 6Karolinska University Hospital, Stockholm, Sweden, 7Karolinska Institutet, Stockholm, Sweden, 8GE Healthcare, Stockholm, Sweden |
||
Fast and high-resolution T2*w and susceptibility-weighted imaging is an essential part of brain MR assessment for many neurological conditions. Very high spatial resolution can be helpful for visualization of microvascular details as well as the central vein sign in white matter Multiple Sclerosis lesions, but such scans have a long acquisition time. In this abstract, we demonstrate high-quality whole-brain T2*w and susceptibility-based MRI at 0.5 mm isotropic resolution in less than 4 minutes using a 3D-echo-planar acquisition and a deep learning reconstruction. Finally, straightforward improvements are discussed for further reduction of the scan time. |
| 2523 | Computer 55
|
Prior-knowledge MRS Metabolite Quantification using Deep Learning Frameworks: A proof-of-concept |
| Federico Turco1 and Johannes Slotboom1 | ||
1SCAN, Institute of Diagnostic and Interventional Neuroradiology, Bern, Switzerland |
||
We implemented a mathematical representation of a prior-knowledge model in a Neural Network using a Tensorflow. The trainables tensors are directly the free parameters of the model and we do metabolite quantification by overfitting the output to the signal that we want to replicate. We found that this way of fitting has a relatively low performance in time domain but similar to the state-of-the-art (TDFDFit) when using frequency domain. In addition we have a faster method and it can be used in future works as a component of a more complex network. |
| 2524 | Computer 56
|
Amide Proton Transfer (APT) Mapping from Undersampled Z-spectra in the Brain Using Deep Learning |
| Karandeep Cheema1,2, Pei Han1,2, Hui Han1, Yibin Xie1,2, Anthony Christodoulou1,2, and Debiao Li1,2 | ||
1Biomedical Imaging Research Institute, Cedars Sinai Medical Center, Los Angeles, CA, United States, 2Department of Bioengineering, University of California, Los Angeles (UCLA), Los Angeles, CA, United States |
||
CEST Imaging requires radio frequency pulses at several frequency offsets to generate CEST maps (APT, NOE, DS, MT). In this study, we aimed to generate APT maps from CEST images of undersampled frequency offsets using deep learning, which can potentially reduce the total scan time of CEST imaging. The Z-spectrum was undersampled by a factor of 3.5 using the Fisher information gain analysis. Fitting results from the proposed method were compared with those from multi-pool fitting with fully sampled Z-spectrum. We have shown that it is feasible to reconstruct APT maps from undersampled, uncorrected CEST images. |
| 2525 | Computer 57
|
Fully Connected Deep Neural Network combined with Segmented Least Square Fitting for improved extraction of IVIM Parameters |
| Alfonso Mastropietro1, Elisa Scalco1, Daniele Procissi2, Nicola Bertolino2, and Giovanna Rizzo1 | ||
1Istituto di Tecnologie Biomediche, Consiglio Nazionale delle Ricerche, Segrate, Italy, 2Radiology, Northwestern University, Chicago, IL, United States |
||
Voxel-by-voxel fitting of intravoxel incoherent motion (IVIM) MRI data using a bi-exponential model is challenging especially with low signal-to-noise ratio (SNR) diffusion-MR images. We propose to combine and use a supervised Deep Neural Network (DNN) approach to increase SNR of the acquired images and thus improve extraction of reliable parameter estimation using a segmented least square fitting algorithm. The effectiveness of the proposed method was demonstrated in both simulated and acquired in vivo data. The proposed approach is promising and can increase performance of the fitting algorithms especially for the case of images with high background noise. |
| 2526 | Computer 58
|
Improved detection of multiple kidney pH compartments by deep learning in MRS and MRSI with hyperpolarized 13C-labelled zymonic acid |
| Martin Grashei1, Wai-Yan Ryana Fok2, Jason G. Skinner1, Bjoern H. Menze2, and Franz Schilling1 | ||
1Department of Nuclear Medicine, TUM School of Medicine, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany, 2Department of Informatics, Technical University of Munich, Munich, Germany |
||
Accurate determination of peak position is challenging for spectra with dense spectral regions paired with low SNR as occuring in pH measurements using hyperpolarized [1,5-13C2,3,6,6,6-D4]zymonic acid in kidney of mice. Despite scarcity of available data from preclinical experiments, convolutional neural networks (CNN) and multilayer perceptrons (MLP) could be trained by complementing real and augmented data with synthetic spectra. While MLPs do not achieve suitable performance, CNNs predict pH compartments with an accuracy comparable or superior to supervised line fitting in synthetic test spectra. Further, CNNs allow generation of composite pH maps with improved quality while quantitatively agreeing with line-fitted maps. |
| 2527 | Computer 59
|
Deep learning framework for accelerating revised-NODDI parameter estimation with tensor-valued diffusion encoding |
| Michele Guerreri1,2, Hojjat Azadbakht2, and Hui Zhang1 | ||
1Computer Science & Centre for Medical Image Computing, University College London, London, United Kingdom, 2AINOSTICS, Manchester, United Kingdom |
||
This work demonstrates the feasibility of using deep learning (DL) to accelerate revised-NODDI parameter estimation with data acquired using tensor-valued diffusion encoding (TVDE). Revised-NODDI is a recently proposed version of NODDI which showed improved compatibility with TVDE. Thanks to this compatibility the model has an extra free parameter to be estimated which, with conventional fitting methods, further slowdown NODDI’s time-demanding parameter estimation. DL methods can vastly accelerate this process. We show that accurate estimation of revised-NODDI parameters can be obtained via a DL framework. We compare the results with those obtained with conventional fitting methods. |
| 2528 | Computer 60
|
Revised-NODDI with conventional dMRI data enabled by deep learning |
| Michele Guerreri1,2, Hojjat Azadbakht2, and Hui Zhang1 | ||
1Computer Science & Centre for Medical Image Computing, University College London, London, United Kingdom, 2AINOSTICS, Manchester, United Kingdom |
||
This work shows that deep learning (DL) enables revised-NODDI parameter estimation from conventional dMRI data. Revised-NODDI is a recently proposed model which overcomes some limitations of NODDI. With conventional fitting methods, revised-NODDI parameters can be robustly estimated only in the presence of data acquired with multiple tensor-valued diffusion encodings. However, this new generation of acquisitions is not yet routinely available in clinical research. We show that revised-NODDI parameters estimated using conventional dMRI data via a DL framework are comparable with the parameters estimated fitting the model to data acquired using multiple tensor-valued diffusion encoding. |
| 2529 | Computer 61
|
DCE-DRONE: Perfusion MRI Parameter Estimation using a DRONE Neural Network |
| Soudabeh Kargar1, Ouri Cohen2, and Ricardo Otazo2 | ||
1MSKCC, NEW YORK, NY, United States, 2MSKCC, New York, NY, United States |
||
In this work, we demonstrate the estimation of DCE acquired perfusion parameters using a DRONE neural network trained on numeric simulated data. Experiments in a digital phantom are used to demonstrate the feasibility of our approach and the clinical utility is shown in subjects with gynecological tumors. |
| 2530 | Computer 62
|
Quantitative MRI parameter estimation with supervised deep learning: MLE-derived labels outperform groundtruth labels |
| Sean Epstein1, Timothy J.P. Bray2, Margaret A. Hall-Craggs2, and Hui Zhang1 | ||
1Centre for Medical Image Computing, University College London, London, United Kingdom, 2University College London, London, United Kingdom |
||
We propose a novel deep learning technique for quantitative MRI parameter estimation. Our method is trained to map noisy qMRI signals to conventional best-fit parameter labels, which act as proxies for the groundtruth parameters we wish to estimate. We show that this training leads to more accurate predictions of groundtruth model parameters than traditional approaches which train on these groundtruths directly. Furthermore, we show that our proposed method is both conceptually and empirically equivalent to existing unsupervisedapproaches, with the advantage of being formulated as supervised training. |
| 2531 | Computer 63
|
SUPER-IVIM-DC, A supervised deep-learning with data consistency approach for IVIM model parameter estimation from Diffusion-Weighted MRI data |
| Elad Rotman1, Onur Afacan2, Sila Kurugol2, Simon K Warfield2, and Moti Freiman1 | ||
1Faculty of Biomedical Engineering, Technion-IIT, Haifa, Israel, 2Computational Radiology lab, Boston Children’s Hospital, Harvard Medical School, Boston, MA, United States |
||
Recently, unsupervised deep-learning showed improved performance in estimating the “Intra-Voxel incoherent motion” (IVIM) signal decay model parameters from Diffusion-weighted Magnetic Resonance Imaging (DW-MRI) data compared to classical methods. However, such deep-learning models do not generalize well on acquisitions with a high signal to noise ratio (SNR). In this work, we introduce SUPER-IVIM-DC, a supervised deep learning network coupled with a data consistency term to improve the capacity of deep-learning-based models to generalize the IVIM signal decay model. We demonstrated an improvement in model generalization, accuracy, and homogeneity using simulation, phantom, and in-vivo experiments. |
| 2532 | Computer 64
|
Investigating complex-valued neural networks applied to phase-cycled bSSFP for multi-parametric quantitative tissue characterization |
| Florian Birk1, Julius Steiglechner1, Klaus Scheffler1,2, and Rahel Heule1,2 | ||
1Max Planck Institute for Biological Cybernetics, Tübingen, Germany, 2Department of Biomedical Magnetic Resonance, University of Tübingen, Tübingen, Germany |
||
The bSSFP sequence is highly sensitive to relaxation parameters, tissue microstructure, and off-resonance frequencies, which has recently been shown to enable multi-parametric tissue characterization in the human brain using real-valued NNs. In this work, a new approach based on complex-valued NNs for voxel-wise simultaneous multi-parametric quantitative mapping with phase-cycled bSSFP input data is presented, possibly facilitating data handling. Relaxometry parameters (T1, T2) and field map estimates (B1+, ΔB0) could be quantified with high robustness and accuracy. The quantitative results were compared for different activation functions, favoring phase-sensitive implementations. |
| 2589 | Computer 30
|
3D Cartesian T1rho Magnetic Resonance Fingerprinting Sequence Design for Evaluation of Cartilage and Skeletal Muscle in the Knee |
| Brendan L. Eck1, Jeehun Kim2, Mingrui Yang2, Dan Ma3, Mark A. Griswold3,4, and Xiaojuan Li2,3 | ||
1Imaging Institute, Cleveland Clinic, Cleveland, OH, United States, 2Biomedical Engineering, Cleveland Clinic, Cleveland, OH, United States, 3Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, 4Department of Radiology, Case Western Reserve University, Cleveland, OH, United States |
||
Magnetic resonance fingerprinting offers a promising framework by which to rapidly quantify T1rho relaxation alongside other tissue properties such as T1 and T2. However, T1rho-MRF has only recently been reported and sequence optimization remains under-explored, particularly for 3D sequences. Using a digital phantom constructed from real knee tissue property maps, we investigate sequence parameters for a 3D Cartesian T1rho-MRF sequence including preparation pulse scheduling and timing, flip angles, number of readouts, and acceleration factor. T1, T2, and T1rho quantification errors in cartilage and skeletal muscle were evaluated. |
| 2590 | Computer 31
|
Preliminary development of a Magnetic Resonance Fingerprinting framework for fast multiparametric low-field NMR relaxometry |
| Giovanni Vito Spinelli1, Leonardo Brizi1,2, Marco Barbieri3, Fabiana Zama4, Germana Landi4, Villiam Bortolotti5, Daniel Remondini1,2, and Claudia Testa1,2 | ||
1Department of Physics and Astronomy "Augusto Righi", University of Bologna, Bologna, Italy, 2INFN, Istituto Nazionale di Fisica Nucleare, Sezione di Bologna, Bologna, Italy, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4Department of Mathematics, University of Bologna, Bologna, Italy, 5Department of Civil, Chemical, Environmental, and Materials Engineering, University of Bologna, Bologna, Italy |
||
A novel application of Magnetic Resonance Fingerprinting (MRF) on low-field NMR is presented. To successfully implement MRF, the correlation between the static and the radio-frequency fields has to be measured, because the evolution of the signal cannot be analyzed without accounting for magnetic fields inhomogeneities. Experimental results have been validated by simulations and compared using the RMSE. This preparatory evaluation allows the use of MRF for NMR parameters quantification. Then, an artificial intelligence approach for parameters reconstruction has been used to overcome the limitation of the standard dictionary approaches when several parameters have to be estimated. |
| 2591 | Computer 32
|
Searching for an MR Fingerprinting sequence to measure brain oxygenation without contrast agent |
| Thomas Coudert1, Aurélien Delphin1, Jan M Warnking1, Benjamin Lemasson1,2, Emmanuel Luc Barbier1, and Thomas Christen1 | ||
1Univ. Grenoble Alpes, Inserm, U1216, Grenoble Institut Neurosciences, GIN, Grenoble, France, 2MoGlimaging Network, HTE Program of the French Cancer Plan, Toulouse, France |
||
We propose to use numerical simulations and Monte Carlo approaches to test 40 MR fingerprinting sequences for their ability to quantify the BOLD effect and provide maps of baseline blood oxygen saturation without the need for contrast agent injection. Our results suggest that several sequences can produce SO2 maps with less that 3% error on average even in the presence of B1 and B0 inhomogeneities. T2 and blood vessel radius could also be estimated with the same acquisitions. |
| 2592 | Computer 33
|
Using 3D realistic blood vessel structures and machine learning for MR vascular Fingerprinting |
| Aurélien Delphin1, Fabien Boux1,2, Clément Brossard1,3, Jan M Warnking1, Benjamin Lemasson1,3, Emmanuel Luc Barbier1, and Thomas Christen1 | ||
1Univ. Grenoble Alpes, Inserm, U1216, Grenoble Institut Neurosciences, GIN, 38000, Grenoble, France, 2Univ. Grenoble Alpes, Inria, CNRS, G-INP, 38000, Grenoble, France, 3MoGlimaging Network, HTE Program of the French Cancer Plan, Toulouse, France |
||
MR vascular fingerprinting aims at mapping cerebral vascular properties such as blood volume and blood oxygenation. We propose to improve the technique by generating dictionaries based on 3D vascular networks segmented from whole brain high-resolution (3 µm isotropic) microscopy datasets. In order to compensate for the limited number of available data and long computing times, we used a machine-learning reconstruction process to generalize our results and tested our approach in healthy, stroke and tumor animal models. Our results show high quality maps with expected contrast and baseline values in healthy animals as well as expected trends in pathological tissues. |
| 2593 | Computer 34
|
Towards A Clinical Prostate MR Fingerprinting Protocol |
| Manuel Baumann1, Jochen Keupp1, Peter Mazurkewitz1, Peter Koken1, Kay Nehrke1, Jakob Meineke1, Thomas Amthor1, and Mariya Doneva1 | ||
1Philips Research Europe, Hamburg, Germany |
||
We present an efficient MR Fingerprinting (MRF) protocol for prostate imaging at 3T. Full prostate coverage is achieved in 04:20 min with no considerable reconstruction latency. Dixon-MRF has been implemented in the scanner software and combined with a B1+ DREAM scan in a fully-integrated B1-corrected MRF reconstruction. Furthermore, flow compensation is included to suppress artifacts caused by blood flow in large vessels. The proposed prostate MRF protocol has been evaluated in six healthy volunteers. |
| 2594 | Computer 35
|
Incorporating saturation bands into MR Fingerprinting reduces streaking artefacts |
| Christopher George Trimble1, Kaia Sørland1, Elise Sandsmark2, Mattijs Elschot1,2, Tone F. Bathen1,2, and Martijn A. Cloos3 | ||
1Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway, 2Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim, Norway, 3Centre for Advanced Imaging, University of Queensland, Brisbane, Australia |
||
Magnetic Resonance Fingerprinting (MRF) enables fast quantitative MR imaging, which is appealing in prostate cancer diagnostics. In pelvic MR Fingerprinting, blood flow in the femoral vessels causes strong streak-like artefacts when a radial sampling strategy is used. Here we develop a strategy to incorporate saturation bands into MRF sequences. Our phantom study shows quantification of areas outside the saturation band are not significantly affected. Meanwhile, qualitative analysis of in vivo experiments indicates a marked reduction in streak intensity when saturation is applied. |
| 2595 | Computer 36
|
Dictionary variance based optimization of MR Fingerprinting |
| Rasim Boyacioglu1, Sherry Huang2, Debra McGivney2, Yong Chen1, and Mark A Griswold1 | ||
1Radiology, Case Western Reserve University, Cleveland, OH, United States, 2Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States |
||
This study presents an optimization tool to investigate and validate the effects of MR sequence modules on MR Fingerprinting. Dictionary variance across the tissue property dimension can predict the outcome of a given change in an MRF sequence. An MRF sequence variant is optimized by varying three typical MRF sequence blocks of FA pattern, RF phase and inversion pulses. The observations from the dictionary variance align with in vivo data results. |
| 2596 | Computer 37
|
Non-Linear Optimization for Enhanced Parameter Retrieval in MR Fingerprinting |
| John T. Lundstrom1,2, Megan E. Poorman3, Andrew Dienstfrey4, and Kathryn E. Keenan3 | ||
1Department of Physics, University of Colorado, Boulder, CO, United States, 2Associate of the National Institute of Standards and Technology, Boulder, CO, United States, 3Physical Measurement Laboratory, National Institute of Standards and Technology, Boulder, CO, United States, 4Information Technology Laboratory, National Institute of Standards and Technology, Boulder, CO, United States |
||
We investigate non-linear optimization to produce quantitative parameter maps in MR Fingerprinting. Our Monte Carlo analysis shows non-linear optimization to be robust with respect to noise. We also find non-linear optimization to yield consistent results when initialized by dictionary matching using either sparse or densely populated dictionaries. Our research outcomes suggest non-linear optimization is an effective enhancement to the MR Fingerprinting pipeline, allowing for smaller dictionary sizes and more accurate parameter retrieval. Future work will include enhancements to, and analysis of, the computational efficiency of non-linear optimization. |
| 2597 | Computer 38
|
Feasibility of Dynamic Contrast-free Vascular Magnetic Resonance Fingerprinting |
| Gregory J. Wheeler1, Quimby N. Lee2, Mary Kate Manhard3, Berkin Bilgic4, and Audrey P. Fan1,2 | ||
1Biomedical Engineering, University of California Davis, Davis, CA, United States, 2Neurology, University of California Davis, Davis, CA, United States, 3Radiology, Cincinnati Children's Hospital, Cincinnati, OH, United States, 4Radiology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States |
||
Vascular magnetic resonance fingerprinting (vMRF) using both magnitude and phase information can achieve reasonable vascular parameter maps of the brain without contrast agents. This approach combined with a rapid, multi-echo pulse sequence enables dynamic vascular function mapping of brain physiology. Here we begin to investigate the feasibility and tradeoffs of performing vMRF with only 5 echoes and a temporal resolution of 5 seconds. Initial findings indicate that reasonable fingerprint matching can be done with only 5 echoes and that the proposed sequence has adequate signal-to-noise ratio for reliable results. |
| 2598 | Computer 39
|
Mitigation of Magnetisation Transfer Effects using Off-Resonance Pulses in MR Fingerprinting |
| Simran Kukran1,2, Iulius Dragonu3, Ben Statton4, Jack Allen5, Pete Lally6, Rebecca Quest7,8, Neal Bangerter7, Dow-Mu Koh9,10, Matthew Orton10,11, and Matthew Grech-Sollars1,12 | ||
1Surgery and Cancer, Imperial College London, London, United Kingdom, 2Radiotherapy and Imaging, Institute of Cancer Research, LONDON, United Kingdom, 3Research and Collaborations GB&I, Siemens Healthcare Ltd, Frimley, United Kingdom, 4London Institute of Medical Sciences,Medical Research Council, Imperial College London, London, United Kingdom, 5National Heart and Lung Institute, Imperial College London, London, United Kingdom, 6Brain Sciences, Imperial College London, London, United Kingdom, 7Department of Bioengineering, Imperial College London, London, United Kingdom, 8Department of Imaging, Imperial College Healthcare NHS Trust, London, United Kingdom, 9Department of Radiotherapy and Imaging, Institute of Cancer Research, London, United Kingdom, 10Department of Radiology, Royal Marsden Hospital, London, United Kingdom, 11Radiotherapy and Imaging, Institute of Cancer Research, London, United Kingdom, 12Department of Medical Physics, Royal Surrey NHS Trust, London, United Kingdom |
||
Previous work has shown an inherent bias in quantitative T1 values for current standard MRF techniques as compared to gold standard methods. In our study we show how magnetisation transfer effects are a component of this bias and that the introduction of off-resonance pulses within an MRF sequence may mitigate these effects. |
| 2599 | Computer 40
|
Accuracy and repeatability of joint sparsity multi-component estimations in MR Fingerprinting |
| Martijn Nagtegaal1, Laura Nunez-Gonzalez2, Dirk Poot2, Jeroen de Bresser3, Thijs van Osch4, Juan Hernandez Tamames2, and Frans Vos1,2 | ||
1Imaging Physics, Delft University of Technology, Delft, Netherlands, 2Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands, 3Department of Radiology, Leiden University Medical Centre, Leiden, Netherlands, 4C.J. Gorter Center for High Field MRI, Radiology Department, Leiden University Medical Centre, Leiden, Netherlands |
||
Multi-component MR Fingerprinting estimations with Sparsity Promoting Iterative Joint Non-negative least squares (SPIJN-MRF) provide the possibility to obtain partial volume tissue segmentations and myelin water maps. We evaluated the accuracy and repeatability of SPIJN-MRF estimations in simulations and 5 subjects, scanned 8 times. The obtained segmentations were compared to segmentations from SPM12 and FSL. In simulations, SPIJN-MRF showed the highest accuracy in total volume and voxel-wise measures. In vivo, higher variation was observed with SPIJN-MRF than with SPM12 and FSL, especially in WM and GM. In conclusion, SPIJN-MRF provides accurate and precise tissue relaxation parameter estimations with partial volume estimations. |
| 2600 | Computer 41
|
Fast distortionless T1-T2 mapping in the prostate using MR fingerprinting with radial acquisition and subspace reconstruction |
| Victoria Y Yu1, Ouri Cohen1, Can Wu1, Ergys Subashi1, Manuel Baumann2, Peter Koken2, Mariya Doneva2, Michael Zelefsky3, Laura Cervino1, and Ricardo Otazo1 | ||
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, 2Philips Research, Hamburg, Germany, 3Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, United States |
||
The application of MR fingerprinting in the prostate is challenging due to the presence of image distortions produced by B0 inhomogeneities. This works presents the combination of radial sampling to minimize the effects of B0 inhomogeneity and temporal subspace reconstruction to accelerate the acquisition for fast distortionless T1 and T2 mapping in the prostate in less than 4 minutes. |
| 2601 | Computer 42
|
Using Machine Learning to Identify Metabolite Spectral Patterns that Reflect Outcome after Cardiac Arrest |
| Marcia Sahaya Louis1,2, Huijun Vicky Liao2, Rohit Singh3, Ajay Joshi1, Jong Woo Lee4, and Alexander Lin2 | ||
1ECE, Boston University, Boston, MA, United States, 2Radiology, Brigham and Women's Hospital, Boston, MA, United States, 3Computer Science and Artificial Intelligence Lab, Massachusetts Institute of Technology, Cambridge, MA, United States, 4Brigham and Women's Hospital, Boston, MA, United States |
||
More than half of patients who undergo targeted temperature management (TTM) after cardiac arrest do not survive hospitalization and 50% of those survivors suffer from long-term cognitive deficits. The goal of this study is to use machine learning methods to characterize the pattern of metabolic changes in patients with good and poor outcomes after cardiac arrest. A machine learning pipeline that incorporates z-scores, decision-tree modeling, principal component analysis, and linear support vector machine was applied to MR spectroscopy data acquired after cardiac arrest. Results confirm that N-acetylaspartate and lactate are important markers but other unexpected findings emerged as well. |
| 2602 | Computer 43
|
Solving T2-blurring: Joint Optimization of Flip Angle Design and DenseNet Parameters for Reduced T2 Blurring in TSE Sequences |
| Hoai Nam Dang1, Jonathan Endres1, Felix Glang2, Simon Weinmüller1, Alexander Loktyushin2, Klaus Scheffler2, Arnd Doerfler1, Andreas Maier3, and Moritz Zaiss1,2 | ||
1Department of Neuroradiology, University Clinic Erlangen, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Erlangen, Germany, 2Magnetic Resonance Center, Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany, 3Department of Computer Science, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Erlangen, Germany |
||
We propose an end-to-end optimization approach to reduce T2 blurring in single-shot TSE sequences, by performing a joint optimization of Flip Angle Design and DenseNet parameters using he original transverse magnetization image at a certain TE as target. Our approach generalizes well to in vivo measurements at 3T, revealing enhanced white and grey matter contrast and fine structures in human brain. |
| 2603 | Computer 44
|
Deep-learning based brain tumor segmentation using quantitative MRI |
| Iulian Emil Tampu1,2, Ida Blystad2,3,4, Neda Haj-Hosseini1, and Anders Eklund1,5 | ||
1Biomedical Engineering, Linkoping University, Linkoping, Sweden, 2Center for Medical Image Science and Visualization (CMIV), Linkoping University, Linkoping, Sweden, 3Department of Radiology in Linköping, Region Östergötland, Center for Diagnostics, Linkoping, Sweden, 4Department of Health, Medicine and Caring Sciences, Division of Diagnostics and Specialist Medicine, Linkoping University, Linkoping, Sweden, 5Department of Computer and Information Science, Linkoping University, Linkoping, Sweden |
||
Manual annotation of gliomas in magnetic resonance (MR) images is a laborious task, and it is impossible to identify active tumor regions not enhanced in the conventionally acquired MR modalities. Recently, quantitative MRI (qMRI) has shown capability in capturing tumor-like values beyond the visible tumor structure. Aiming at addressing the challenges of manual annotation, qMRI data was used to train a 2D U-Net deep-learning model for brain tumor segmentation. Results on the available data show that a 7% higher Dice score is obtained when training the model on qMRI post-contrast images compared to when the conventional MR images are used. |
| 2604 | Computer 45
|
Quantitative Radiomic Features of Deep Learning Image Reconstruction in MRI |
| Edward J Peake1, Andy N Priest1,2, and Martin J Graves1,2 | ||
1Imaging, Cambridge University Hospitals NHS Foundation Trust, Cambridge, United Kingdom, 2Department of Radiology, University of Cambridge, Cambridge, United Kingdom |
||
Radiomic features are sensitive to changes in imaging parameters in MRI. This makes it challenging to develop robust machine learning models using imaging features. We explore the effect of clinically available deep learning image reconstruction on the performance of radiomic features. Correlation coefficient values varied (0.56 - 1.00) when comparing radiomic features of deep learning reconstructed images and ‘conventional’ MRI scans. The noise reduction level had a large impact on correlation coefficients, but variations were also significant between different types of imaging feature. Identification of highly correlated features may help identify more stable sets of radiomic features for machine learning. |
| 2605 | Computer 46
|
Spherical-CNN based diffusion MRI parameter estimation is robust to gradient schemes and equivariant to rotation |
| Tobias Goodwin-Allcock1, Robert Gray2, Parashkev Nachev2, Jason McEwan3, and Hui Zhang1 | ||
1Department of Computer Science and Centre for Medical Image Computing, UCL, London, United Kingdom, 2Department of Brain Repair & Rehabilitation, Institute of Neurology, UCL, London, United Kingdom, 3Kagenova Limited, Guildford, United Kingdom |
||
We demonstrate the advantages of spherical convolutional neural networks over conventional fully connected networks at estimating rotationally invariant microstructure indices. Fully-connected networks (FCN) have outperformed conventional model fitting for estimating microstructure indices, such as FA. However, these methods are not robust to changes diffusion weighted image sampling scheme nor are they rotationally equivariant. Recently spherical-CNN have been supposed as a solution to this problem. However, the advantages of spherical-CNNs have not been leveraged. We demonstrate both spherical-CNNs robust to new gradient schemes as well as the rotational equivariance. This has potential to decrease the number of training datapoints required. |
| 2606 | Computer 47
|
Comparison of machine learning methods for detection of prostate cancer using bpMRI radiomics features |
| Ethan J Ulrich1, Jasser Dhaouadi1, Robben Schat2, Benjamin Spilseth2, and Randall Jones1 | ||
1Bot Image, Omaha, NE, United States, 2Radiology, University of Minnesota, Minneapolis, MN, United States |
||
Multiple prostate cancer detection AI models—including random forest, neural network, XGBoost, and a novel boosted parallel random forest (bpRF)—are trained and tested using radiomics features from 958 bi-parametric MRI (bpMRI) studies from 5 different MRI platforms. After data preprocessing—consisting of prostate segmentation, registration, and intensity normalization—radiomic features are extracted from the images at the pixel level. The AI models are evaluated using 5-fold cross-validation for their ability to detect and classify cancerous prostate lesions. The free-response ROC (FROC) analysis demonstrates the superior performance of the bpRF model at detecting prostate cancer and reducing false positives. |
| 2607 | Computer 48
|
Isolating cardiac-related pulsatility in blood oxygenation level-dependent MRI with deep learning |
| Jake Valsamis1, Nicholas Luciw1, Nandinee Haq1, Sarah Atwi 1, Simon Duchesne 2,3, William Cameron1, and Bradley J MacIntosh1,4 | ||
1Physical Sciences, Sunnybrook Research Institute, Toronto, ON, Canada, 2Radiology Department, Faculty of Medicine, Laval University, Québec, QC, Canada, 3Quebec City Mental Health Institute, Québec, QC, Canada, 4Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada |
||
Persistent exposure to highly pulsatile blood can damage the brain’s microvasculature. A convenient method for measuring cerebral pulsatility would allow investigation into its relationship with vascular dysfunction and cognitive decline. In this work, we propose a convolutional neural network (CNN) based deep learning solution to estimate cerebral pulsatility using only the frequency content from BOLD MRI scans. Various frequency component inputs were assessed, and echo time dependence was evaluated with a 5-fold cross-validation. Pulsatility was estimated from BOLD MRI data acquired on a different scanner to assess generalizability. The CNN reliably estimated pulsatility and was robust to various scan parameters. |
| 2608 | Computer 49
|
Image enhancement of Quasi-Diffusion Imaging using a fully connected neural network |
| Ian Robert Storey1, Catherine Anne Spilling1,2, Xujiong Ye3, Thomas Richard Barrick1, and Franklyn Arron Howe1 | ||
1St George's, University of London, London, United Kingdom, 2King's College London, London, United Kingdom, 3University of Lincoln, Lincoln, United Kingdom |
||
A fully connected neural network (FCN) was trained to map a short diffusion weighted image acquisition to high quality Quasi-Diffusion Imaging (QDI) parameter maps. The FCN produced denoised and enhanced QDI parameter maps compared to weighted least squares fitting of data to the QDI model. The FCN shows generalisation to unseen pathology such as grade IV glioma dMRI data and demonstrates the FCN can produce high quality QDI tensor maps from clinically feasible 2 minute data acquisitions. An FCN further enhances the ability of QDI to provide non-Gaussian diffusion imaging within clinically feasible acquisition times. |
| 2609 | Computer 50
|
Hybrid supervised and self-supervised deep learning for quantitative mapping from weighted images using low-resolution labels |
| Shihan Qiu1,2, Anthony G. Christodoulou1,2, Yibin Xie1, and Debiao Li1,2 | ||
1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, 2Department of Bioengineering, UCLA, Los Angeles, CA, United States |
||
Deep learning methods have been developed to estimate quantitative maps from conventional weighted images, which has the potential to improve the availability and clinical impact of quantitative MRI. However, high-resolution labels required for network training are not commonly available in practice. In this work, a hybrid supervised and physics-informed self-supervised loss function was proposed to train parameter estimation networks when only limited low-resolution labels are accessible. By taking advantage of high-resolution information from the input weighted images, the proposed method generated sharp quantitative maps and had improved performance over the supervised training method purely relying on the low-resolution labels. |
| 2840 | Computer 89
|
Removal of perfusion lag structure before ICA-FIX cleaning boosts both reproducibility and between-subject similarity in fcMRI |
| Toshihiko Aso1 and Takuya Hayashi1 | ||
1Laboratory for Brain Connectomics Imaging, RIKEN Center for Biosystems Dynamics Research, Kobe, Japan |
||
Removal of the perfusion lag structure in the fMRI signal was combined with independent component analysis (ICA)-based cleaning to evaluate its effect on test-retest reproducibility. In HCP test-retest data from 40 subjects, either treatment and the combined method effectively improved the reproducibility of functional connectivity. However, the effect was more evident in between-subject similarity of the connectivity pattern than within-subject similarity. This finding suggests that the perfusion lag structure may be one of the sources of between-subject variability, that may degrade the functional connectivity. |
| 2841 | Computer 90
|
Low-grade glioma resting-state fMRI pre-operative maps: a comparison with functional atlas from direct electrode stimulation. |
| Beatrice Federica Luciani1, Donna Gift Cabalo1, Francesca Saviola1, Luca Zigiotto2,3, Stefano Tambalo1, Domenico Zacà1, Lisa Novello1, Silvio Sarubbo2,3, and Jorge Jovicich1 | ||
1CIMeC Center for Mind/Brain Sciences, Trento, Italy, 2Department of Neuroscience, Division of Neurosurgery, S.Chiara Hospital, APSS Trento, Trento, Italy, 3Structural and Functional Connectivity Lab, S.Chiara Hospital, APSS Trento, Trento, Italy |
||
A cortical-subcortical functional atlas has been recently proposed from direct-electrode stimulation on low grade glioma patients. The agreement between that atlas and non-invasive pre-operative resting-state functional MRI (rs-fMRI) data has not yet been investigated. In this study, we use pre-operative rs-fMRI data and the RestNeuMap tool to investigate the agreement between language and sensory networks in a group of low grade glioma patients. We find that rs-fMRI networks of language and speech arrest show greater overlap with DES-derived functional atlas than the sensorimotor network. |
| 2842 | Computer 91
|
Mitigation of noise floor in diffusion MRI using deep learning |
| Hu Cheng1, Jian Wang2, Sophia Vinci-Booher1, Qiuting Wen3, and Sharlene Newman4 | ||
1Psychological and Brain Sciences, Indiana University, Bloomington, IN, United States, 2Shandong Normal University, Jinan, China, 3Indiana University, Indianapolis, IN, United States, 4University of Alabama, Tuscaloosa, AL, United States |
||
Rectified noise floor is a challenging problem for high-resolution diffusion MRI, which cannot be tackled by current denoising methods. We propose a simple deep learning method for correcting noise floor in diffusion MRI. The method is based on 1D CNN model and works on voxel-wise time courses. Therefore, even one dataset of the brain has sufficient number of samples for training, which is a big advantage for practical application. Both simulation and in vivo results show that the method is robust in mitigating the noise floor artifact and restore the true values of diffusion metrics. |
| 2843 | Computer 92
|
Distortion correction of 3D EPI for cortical depth-resolved fMRI at 7T |
| Vahid Malekian1, Nadine N. Graedel1, Alice L. Hickling1, Ali Aghaeifar1,2, Nadège Corbin1,3, Oliver Josephs1, Eleanor A. Maguire1, and Martina F. Callaghan1 | ||
1Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 2MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom, 3Centre de Résonance Magnétique des Systèmes Biologiques, CNRS‐University Bordeaux, Bordeaux, France |
||
Geometric distortion is a major concern for cortical depth-dependent fMRI using EPI readouts that suffer from low bandwidth in the phase-encoded direction. Distortions are exacerbated at higher field strengths due to increased B0 field inhomogeneity. In this study, we quantitatively compared two distortion correction methods (B0 field-mapping and reversed-PE) applied to submillimeter (0.8mm isotropic) 3DEPI data at 7T. Cortical alignment was evaluated through comparison with an anatomically-faithful MP2RAGE reference by computing the dice coefficient and normalised mutual information. Both distortion correction methods usefully improve alignment. The reversed-PE approach performed better.
|
| 2844 | Computer 93
|
Monte Carlo simulations of weighted overlap map thresholds to reduce the risk for type I errors in fMRI |
| Peter Van Schuerbeek1 | ||
1Radiology, UZ Brussel (VUB), Brussels, Belgium |
||
We performed Monte Carlo Simulations to show that the risk for type I errors decreases if a minimum overlap between the individual results is set in addition to the significance of the group results. The activation maps of a real fMRI experiment showed that the combination of a weighted overlap map threshold in combination with a voxel significance threshold seems to lead to less type II errors compared to applying a family wise error (FWE) correction. |
| 2845 | Computer 94
|
Adaptive non-local means filtering as a drop-in preprocessing step to improve statistical sensitivity in task-based fMRI |
| Ajay Nemani1 and Mark J Lowe2 | ||
1Imaging Institute, Cleveland Clinic, Cleveland, OH, United States, 2Cleveland Clinic, Cleveland, OH, United States |
||
Spatial filtering is an important step in the preprocessing of task-based fMRI to improve sensitivity in statistical analyses. This is usually implemented as a pure distance-based filter such as Gaussian filtering or an optimized matched filter. Adaptive non-local means (ANLM) filtering is a patch-based approach that is sensitive to the local noise model, especially at low signal to noise ratio such as fMRI. We show how ANLM filtering is a simple drop-in replacement at the spatial smoothing step of fMRI preprocessing pipeline that compares favorably to other approaches while better preserving local high frequency features. |
| 2846 | Computer 95
|
NODDI-FAST derived parameters are echo-time independent |
| Maryam H Alsameen1, Zhaoyuan Gong1, Matthew Kiely1, Curtis Triebswetter1, and Mustapha Bouhrara1 | ||
1Magnetic Resonance Physics of Aging and Dementia Unit, National Institute on Aging, NIH, Baltimore, MD, United States |
||
In a previous work, we introduced the NODDI-FAST approach to address issues regarding overestimated CSF and neurite density (NDI) fractions in white matter seen with the original NODDI approach. However, in both NODDI-FAST and NODDI signal models, the compartment-specific T2 relaxations are not considered. Therefore, derived parameter estimates, especially NDI, could be dependent on echo time (TE). Here, we show that, as expected, ODI derived values using either NODDI or NODDI-FAST are TE-independent. We also confirm that NDI derived values using NODDI are TE-dependent. More importantly, we show that NDI derived values using NODDI-FAST are TE-independent.
|
| 2847 | Computer 96
|
Fractal-Based Analysis of fMRI BOLD Signal During Naturalistic Viewing Conditions |
| Olivia Campbell1,2, Tamara Vanderwal2,3, and Alexander Mark Weber1,2,4,5 | ||
1School of Biomedical Engineering, University of British Columbia, Vancouver, BC, Canada, 2BC Children's Hospital Research Institute, Vancouver, BC, Canada, 3Department of Psychiatry, University of British Columbia, Vancouver, BC, Canada, 4Department of Pediatrics, University of British Columbia, Vancouver, BC, Canada, 5Department of Neuroscience, University of British Columbia, Vancouver, BC, Canada |
||
We investigated the difference between the brain’s fractal dynamics during movie-watching and resting-state conditions using 7T fMRI data. During movie-watching, we find that the BOLD signal becomes more scale-invariant and self-similar than during the resting-state. This supports the idea that movie-watching evokes a state of optimal neural functioning that may better reflect the endogenous state of the brain than a fixed cross-hair. We also find that fractal properties differ greatly across functional networks, providing novel information about the temporal dynamics of each network during naturalistic processing. Overall, these findings advance understanding of both fractal dynamics and naturalistic viewing in fMRI. |
| 2848 | Computer 97
|
Ultra-fast 3D fMRI to explore cardiac-induced fluctuations in BOLD-based functional imaging |
| Brad Sutton1,2,3, Aaron Anderson1, Benjamin Zimmerman1, Paul Camacho1,3,4, Riwei Jin1,2, Charles Marchini1,2, Olawale Salaudeen1, Natalie Ramsy1,5, Davide Boido6, Serge Charpak7, Andrew Webb8, and Luisa Ciobanu6,9 | ||
1Beckman Institute, University of Illinois Urbana Champaign, Urbana, IL, United States, 2Bioengineering, University of Illinois Urbana Champaign, Urbana, IL, United States, 3Neuroscience Program, University of Illinois Urbana Champaign, Urbana, IL, United States, 4Interdisciplinary Health Sciences Institute, University of Illinois Urbana Champaign, Urbana, IL, United States, 5Carle-Illinois College of Medicine, Urbana, IL, United States, 6NeuroSpin CEA, Saclay, France, 7INSERM, Paris, France, 8Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, 9Paris-Saclay University, Saclay, France |
||
Leveraging recent developments in high speed imaging, we use a 3D fMRI acquisition with 60 ms TR to examine the impact of physiological signals on spatiotemporal correlations in BOLD imaging. Given the standard sampling of fMRI with 2D slices, coherent slice-by-slice sampling of physiological signals can lead to many additional components in the fMRI time series. The current approach provides a way to explore the impact of these confounding signals and to test correction methods. |
| 2849 | Computer 98
|
Principal Component Characterization of Deformation Variations Using Dynamic Imaging Atlases |
| Fangxu Xing1, Riwei Jin2, Imani Gilbert3, Georges El Fakhri1, Jamie Perry3, Bradley Sutton2, and Jonghye Woo1 | ||
1Radiology, Massachusetts General Hospital, Boston, MA, United States, 2University of Illinois at Urbana-Champaign, Champaign, IL, United States, 3East Carolina University, Greenville, NC, United States |
||
High-speed dynamic magnetic resonance imaging is a highly efficient tool in capturing vocal tract deformation during speech. However, automated quantification of variations in motion patterns during production of different utterances has been a challenging task due to spatial and temporal misalignments between different image datasets. We present a principal component analysis-based deformation characterization technique built on top of established dynamic speech imaging atlases. Two layers of principal components are extracted to represent common motion and utterance-specific motion, respectively. Comparison between two speech tasks with and without nasalization reveals subtle differences on velopharyngeal deformation reflected in the utterance-specific principal components. |
| 2850 | Computer 99
|
Generalizable synthetic multi-contrast MRI generation using physics-informed convolutional networks |
| Luuk Jacobs1,2, Stefano Mandija1,2, Hongyan Liu1,2, Cornelis AT van den Berg1,2, Alessandro Sbrizzi1,2, and Matteo Maspero1,2 | ||
1Department of Radiotherapy, Division of Imaging and Oncology, UMC Utrecht, Utrecht, Netherlands, 2Computational Imaging Group for MR Diagnostics and Therapy, Center for Image Sciences, UMC Utrecht, Utrecht, Netherlands |
||
Synthetic MRI aims to reconstruct multiple MRI contrasts from short measurements of tissue properties. Here, a generalizable physics-informed deep learning-based approach for synthetic MRI was investigated. Acquired data were mapped to effective quantitative parameter maps, here named q*-maps, which are fed to a physical signal model synthesizing four contrasts-weighted images. We demonstrated that from q*-maps, MRI contrasts unseen during training could be synthesized. The proposed method is benchmarked to a standard end-to-end deep learning approach. The two deep learning methods generated similar brain images for healthy subjects and patients with different pathologies. |
| 2899 | Computer 105
|
Investigating the influences of spectral lineshape and linewidth in LCModel basis-sets on 1H MRS quantification at 7T |
| Ying Xiao1,2, Ivan Tkáč3, and Lijing Xin1 | ||
1Animal imaging and technology (CIBM), EPFL, Lausanne, Switzerland, 2Laboratory for Functional and Metabolic Imaging (LIFMET), EPFL, Lausanne, Switzerland, 3Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States |
||
In this study, we investigated the 1H MR spectra quantification accuracy using basis sets with different linewidths and lineshapes in LCModel. Simulation results showed that the estimated metabolite concentrations depend on the linewidths and lineshapes of the basis sets as well as the spectral lineshape. This effect was validated with in vivo MR results. We conclude that the spectral lineshape and linewidth of basis sets should be carefully considered for accurate metabolite concentration estimations from short-TE 1H MR spectra using LCModel. |
| 2900 | Computer 106
|
Improving SNR in deuterium metabolic imaging: A comparison of methods |
| Elton Tadeu Montrazi1, Ricardo Martinho1, Qingjia Bao2, Keren Sasson1, Lilach Agemy1, Avigdor Scherz1, and Lucio Frydman1 | ||
1Weizmann Institute of Science, Rehovot, Israel, 2Chinese Academy of Sciences, Wuhan, China |
||
Deuterium metabolic imaging (DMI) is a promising approach to study tumor metabolism. Still, DMI’s signal-to-noise ratio (SNR) is limited because of 2H’s low Larmor frequency, coupled to the low concentrations of DMI’s targets. We recently proposed a multi-echo balanced steady-state free precession (ME-bSSFP) method increasing DMI’s SNR, but still find it lacking in some aspects. This work assesses simple apodization, Compressed Sensing Multiplicative (CoSeM) and Block-matching/3D filtering (BM3D) denoising methods for improving DMI’s results. The ability of the latter denoising methods to enhance sensitivity without blurring resolution, is evidenced by pancreatic cancer studies carried at 15.2T. |
| 2901 | Computer 107
|
Phosphorus MR Spectroscopy of Healthy Human Spleen |
| Jan Weis1, Maysam Jafar2, and Per Liss3 | ||
1Department of Medical Physics, Uppsala University Hospital, Uppsala, Sweden, 2Philips Nordic, Stockholm, Sweden, 3Department of Surgical Sciences, Section of Radiology, Uppsala University Hospital, Uppsala, Sweden |
||
Phosphorous spectra of healthy spleen are useful for studies of splenic malignancies and benign causes of splenomegaly. The main challenge is the relatively small size of the normal spleen and the large distance between human spleen and surface coil. The purpose of this study was to investigate whether it is possible to acquire phosphorous spectra of healthy spleen using single-voxel ISIS sequence on a 3T scanner. We demonstrate that the proposed spectroscopy of human spleen is feasible in a clinically acceptable acquisition time and that transmitter excitation profile and chemical shift displacement errors need to be taken into consideration |
| 2902 | Computer 108
|
MP-PCA and Low-rank noise-reduction in 1H-FID-MRSI data in the rat brain at 14.1T |
| Brayan Alves1,2, Dunja Simicic1,2,3, Jessie Mosso1,2,3, Ileana Jelescu4, Cristina Cudalbu1,2, and Antoine Klauser1,5 | ||
1CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 2Animal Imaging and Technology, EPFL, Lausanne, Switzerland, 3LIFMET, EPFL, Lausanne, Switzerland, 4Service de radiodiagnostic et radiologie interventionnelle, Lausanne University Hospital CHUV, Lausanne, Switzerland, 5Department of Radiology and Medical, Informatics, University of Geneva, Geneva, Switzerland |
||
1H-MRSI is highly challenging and the constant appetite for higher spatial resolution leads to increased search for post-processing methods aiming to reduce the noise variance in 1H-MRSI. The aim of the present study was to implement and test the feasibility of two noise-reduction techniques on preclinical 14.1T fast 1H-FID-MRSI datasets: the MP-PCA based denoising and the low-rank TGV reconstruction. Our results are promising showing an enormous potential of the two noise-reduction techniques towards novel and fast MRSI developments. Further studies will be performed to evaluate if the “apparent” increase in spectral SNR translates in true lower uncertainty in metabolite concentrations. |
| 2903 | Computer 109
|
A multiplicative denoising method based on compressed sensing for magnetic resonance spectroscopy and imaging |
| Ricardo P. Martinho1, David Koprivica1, Mihajlo Novakovic1, Michael J. Jaroszewicz1, and Lucio Frydman1 | ||
1Department of Chemical and Biological Physics, Weizmann Institute of Science, Rehovot, Israel |
||
2D MRI/S are commonly affected by changes in temperature, field drifts, or motions, leading to multiplicative noise. We introduce here CoSeM (Compressed Sensing Multiplicative) denoising, a method that converts instability-related “t1” noise, into additive noise liable to signal averaging. CoSeM evaluates and discards indirect-domain points that may have been strongly influenced by instabilities, and makes up for this discarding by compressed sensing reconstructions. 2D localized MRS in the brain and 2D abdominal MRI experiments liable to instabilities, evidence 2-3 fold increases in SNR by CoSeM processing. CoSeM is also shown to retain quantitative information –e.g., in T1 mapping experiments. |
| 2904 | Computer 110
|
Towards a robust and reproducible GluCEST analysis pipeline |
| Joelle Jee1, Ravi Prakash Reddy Nanga2, Abigail Cember2, Paul Jacobs2, Heather Robinson1, Mark A Elliott2, Ravinder Reddy2, and David R Roalf1 | ||
1Psychiatry, University of Pennsylvania, Philadelphia, PA, United States, 2Radiology, University of Pennsylvania, Philadelphia, PA, United States |
||
Robust, reproducible methods are paramount in neuroimaging. This is a challenge for leading-edge and novel techniques at 7T MRI, such as glutamate-weighted chemical exchange saturation transfer (GluCEST). As such, our primary objective is to move towards a robust and reproducible analytical pipeline for GluCEST imaging data. Here we show, using python-based neuroimaging tools, the development of GluCEST-prep. GluCEST-prep incorporates common neuroimaging analysis steps—brain extraction, tissue segmentation, co-registration and normalization—that are optimized for 7T MRI and uses python-based analysis—in place of in-house MATLAB scripts—to generate post-processed 2D or 3D GluCEST images in both native and template space. |
| 2905 | Computer 111
|
Improved MP-PCA denoising through tensor generalization |
| Jonas Lynge Olesen1,2, Andrada Ianus3, Noam Shemesh3, and Sune Nørhøj Jespersen1,2 | ||
1Center of Functionally Integrative Neuroscience (CFIN) and MINDLab, Aarhus University, Aarhus, Denmark, 2Department of Physics and Astronomy, Aarhus University, Aarhus, Denmark, 3Champalimaud Research, Champalimaud Centre for the Unknown, Lisbon, Portugal |
||
A popular SNR-boosting method in MRI is denoising based on principal component analysis with automated rank estimation by exploiting the Marchenko-Pastur distribution of noise singular values (MP-PCA). MP-PCA operates by reshaping data-patches into matrices to discriminate signal from noise using random matrix theory. Here, we generalize MP-PCA to exploit tensor-structured data arising in, e.g., multi-contrast or multicoil acquisitions, without introducing new assumptions. As proof of concept, we demonstrate a substantial increase in denoising performance in a multi-TE DKI dataset, in particular for small sliding windows. This is beneficial especially in cases of rapidly varying contrast or spatially varying noise. |
| 2906 | Computer 112
|
Constructing a Mesh-Based Template of the Liver is Consistent as a Function of Cohort Size in the UK Biobank |
| Marjola Thanaj1, Nicolas Basty1, Madeleine Cule2, Elena P Sorokin2, Jimmy David Bell1, Elizabeth Louise Thomas1, and Brandon Whitcher1 | ||
1Research Centre for Optimal Health, School of Life Science, University of Westminster, London, United Kingdom, 2Calico Life Sciences LLC, South San Francisco, CA, United States |
||
Computational atlases provide many advantages in quantifying and modeling differences in the size and shape of internal organs between individuals in population imaging studies. A cohort of UK Biobank participants were analyzed to explore the consistency of templates, which is the average labeled image from a number of subjects constructed across different population sizes, and investigate whether this has an impact on the statistical analysis. We found no differences in our significance areas between templates. Although, for the statistical parametric mapping, a larger sample size that involves multiple hypothesis tests, results in a higher representative power. |
| 2907 | Computer 113
|
Evaluating Reproducibility and Repeatability of Penalty Function Based Methods for Quantitative Intravoxel Incoherent Motion Analysis |
| Esha Baidya Kayal1, Kedar Khare1, Raju Sharma2, Sameer Bakhshi2, Devasenathipathy Kandasamy2, and Amit Mehndiratta1,2 | ||
1Indian Institute of Technology Delhi, Delhi, India, 2All India Institute of Medical Sciences Delhi, Delhi, India |
||
Reproducibility and repeatability of penalty-based IVIM analysis methodologies BE+TV and BE+HPF have been evaluated using clinical dataset with osteosarcoma in comparison with existing IVIM analysis methods. IVIM datasets at three time-points during chemotherapy were analyzed and healthy muscle tissue was used as local control for IVIM analysis and evaluation of scan-rescan repeatability and reproducibility. Results showed, parametric maps estimated by BE+TV and BE+HPF methods were observed to have higher reproducibility and lower spatial inhomogeneity than the existing IVIM analysis methods and produced satisfactory inter-scan agreement in parameters estimation proving its repeatability in clinical scenario. |
| 2908 | Computer 114
|
Comparison of intra-voxel spatial distributions of diffusion coefficients with intra-voxel spatial distributions of kurtosis coefficients |
| Suguru Yokosawa1, Toru Shirai1, Yoshitaka Bito2, and Hisaaki Ochi1 | ||
1Innovative Technology Laboratory, FUJIFILM Healthcare Corporation, Tokyo, Japan, 2Radiation Diagnostic Systems Division, FUJIFILM Healthcare Corporation, Chiba, Japan |
||
Previously, we have proposed the method for characterizing the intra-voxel spatial distribution of apparent diffusion coefficients (ADC) by using texture analysis and showed that texture features can provide different information from the conventional diffusion tensor imaging. In this study, we extended our method to the apparent kurtosis coefficients (AKC). We compared the features of the intra-voxel spatial distribution of ADC with that of AKC. We found that there is no significant difference in the spatial distribution of AKC from that of ADC, and a short scanning time with a single b-value may be sufficient to obtain information on diffusion distribution. |
| 2909 | Computer 115
|
Disentangling apparent discordance between ASL-MRI and [18F]-FDG PET following a single dose of the β2-agonist clenbuterol |
| Courtney A. Bishop1, Gaia Rizzo1, Thomas Lodeweyckx2, Jan de Hoon2, Koen Van Laere3,4, Michel Koole4, Wim Vandenberghe5, Eugenii Rabiner1,6, Renee Martin7, Anthony Ford7, and Gabriel Vargas7 | ||
1Invicro, London, United Kingdom, 2Center for Clinical Pharmacology, Department of Pharmaceutical and Pharmacological Sciences, KU Leuven, Leuven, Belgium, 3Division of Nuclear Medicine, University Hospital Leuven, Leuven, Belgium, 4Nuclear Medicine and Molecular Imaging, KU Leuven, Leuven, Belgium, 5Department of Neurology, University Hospital Leuven, Leuven, Belgium, 6Centre for Neuroimaging Sciences, Institute of Psychiatry, Psychology and Neuroscience, King’s College, London, United Kingdom, 7CuraSen Therapeutics, San Carlos, CA, United States |
||
Here we demonstrate that an apparent discrepancy in CBF (from ASL) and CMRglu (from [18F]-FDG PET) following high-dose administration of clenbuterol in healthy volunteers is caused by within-scan rising plasma glucose concentrations. With simulation-based correction, post-clenbuterol CBF changes from ASL agree with blood flow estimates from [18F]-FDG PET, without increase in CMRGlu. ASL-MRI may therefore provide a valuable tool for monitoring the central effects of β2-adrenergic receptor activation in larger, future studies on both healthy volunteers and patients with neurodegenerative disorders. |
| 2682 | Computer 46
|
SANDIAMICO: an open-source toolbox for Soma And Neurite Density Imaging (SANDI) with AMICO |
| Simona Schiavi1,2, Mario Ocampo-Pineda1, Raphaël Truffet3, Emmanuel Caruyer3, Alessandro Daducci1, and Marco Palombo4,5,6 | ||
1Department of Computer Science, University of Verona, Verona, Italy, 2Department of Neuroscience, Rehabilitation, Ophthalmology, Genetics, Maternal and Child Health (DINOGMI), University of Genoa, Genoa, Italy, 3CNRS, Inria, Inserm, Univ Rennes, Rennes, France, 4Centre for Medical Image Computing (CMIC), Department of Computer Science, University College London, London, United Kingdom, 5Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, United Kingdom, 6School of Computer Science and Informatics, Cardiff University, Cardiff, United Kingdom |
||
The soma and neurite density imaging (SANDI) model has been recently introduced to estimate diffusion MRI indices of apparent neurite and soma density noninvasively in the brain. Here, we introduce SANDIAMICO a fast (10 seconds for whole brain images) and robust implementation for estimating SANDI parameters using the Accelerated Microstructure Imaging via Convex Optimization (AMICO) framework. Using numerical simulations and in vivo human data, we show excellent performances of SANDIAMICO in terms of accuracy, precision, robustness to noise and intra-subject reproducibility. |
| 2683 | Computer 47
|
Quality control and nuisance regression of fMRI, looking out where signal should not be found |
| Céline Provins1, Christopher J. Markiewicz2, Rastko Ciric2, Mathias Goncalves2, Cesar Caballero-Gaudes3, Russell A. Poldrack2, Patric Hagmann1, and Oscar Esteban1 | ||
1Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 2Department of Psychology, Stanford University, Stanford, CA, US, Stanford, CA, United States, 3Basque Center on Cognition, Brain and Language, Donostia, Spain, Donostia, Spain |
||
Quality control of functional MRI data is essential as artifacts can have a critical impact on subsequent analysis. Yet, visual assessment of a dataset is tedious and time-consuming. By extending the carpet plot with the voxels located on a closed band (or “crown”) around the brain, we showed that fMRI data quality can be assessed more effectively. This new feature has been incorporated into MRIQC and fMRIPrep. In addition, a new nuisance regressor has been added to the latter, calculated from timeseries within this new “crown”. |
| 2684 | Computer 48
|
Data-driven heatmap clustering approach to analyzing structural connectomes |
| Wilburn E Reddick1, Jordan Teague1, Ruitian Song1, John O Glass1, and Lisa M Jacola2 | ||
1Diagnostic Imaging, St. Jude Children's Research Hospital, Memphis, TN, United States, 2Psychology, St. Jude Children's Research Hospital, Memphis, TN, United States |
||
Heatmap differences between 336 patients treated for ALL and 59 age-similar controls were calculated for a limited set of reproducible edges and clustered using an agglomerative hierarchical clustering that identified four clusters with unique connectivity. Differences in edges demonstrated patterns of lower (group 2), higher (group 4), or mixed (group 1 & 3) connectivity relative to controls. Neurocognitive performance was substantially below normal on measures of memory (DSF – groups 2 & 3; DSB - group 1; Visual matching – group 2). This data driven approach was able to identify four distinct groups of patients with unique connectivity profiles and cognitive performance. |
| 2685 | Computer 49
|
Acquisition and finite element analysis of 4DFlow-MR functional angiography in brain and neck |
| María Paula Del Pópolo1,2, Leandro Enrique Salcedo1,2,3, Clara Lisazo1,2, Federico Julián González Nicolini2,4,5, Trinidad González Padin1,2, Rodrigo Nahuel Alcalá Marañón1,2,6, Rocío Paula Boschi2,7, Mercedes Paula Caspi2,7, Ezequiel F Petra8, Pedro Pablo Ariza1,2, Valdir Fialkowski9, and Daniel Fino Villamil1,2,5 | ||
1MRI Department, Fundación Argentina para el Desarrollo en Salud, Mendoza, Argentina, 2MRI Department, Fundación Escuela de Medicina Nuclear, Mendoza, Argentina, 3FI, Universidad de Mendoza, Mendoza, Argentina, 4FGQNYCS, Comisión Nacional de Energía Atómica, Buenos Aires, Argentina, 5Instituto Balseiro, Universidad Nacional de Cuyo, San Carlos de Bariloche, Argentina, 6FCEN, Universidad Nacional de Cuyo, Mendoza, Argentina, 7MRI Department, Fundación Argentina para el Desarrollo en Salud, Mendoza, Argentina, 8Hemodynamics, Fundación Argentina para el Desarrollo en Salud, Mendoza, Argentina, 9Research Department, Philips Healthcare, Barueri/San Paulo, Brazil |
||
The abscence of an angiographic functional analysis of the head and neck regions opens the possibility of applying the 4D Flow technique to quantify hemodynamics and blood flow parameters. Through a Python algorithm using Finite Elements Method the dynamic blood behavior was analyzed, obtaining a complete correlation with the clinical diagnosis. |
| 2686 | Computer 50
|
SynthStrip: skull stripping for any brain image |
| Andrew Hoopes1, Jocelyn S. Mora1, Adrian Dalca1,2,3, Bruce Fischl*1,2,3, and Malte Hoffmann*1,2 | ||
1Martinos Center for Biomedical Imaging, Boston, MA, United States, 2Department of Radiology, Harvard Medical School, Boston, MA, United States, 3Computer Science and Artificial Intelligence Lab, Massachusetts Institute of Technology, Cambridge, MA, United States |
||
The removal of non-brain signal from MR data is an integral component of neuroimaging streams. However, popular skull-stripping utilities are typically tailored to isotropic T1-weighted scans and tend to fail, sometimes catastrophically, on images with other MRI contrasts or stack-of-slices acquisitions that are common in the clinic. We propose SynthStrip, a flexible tool that produces highly accurate brain masks across a landscape of neuroimaging data with widely varying contrast and resolution. We implement our method by leveraging anatomical label maps to synthesize a broad set of training images, optimizing a robust convolutional network agnostic to MRI contrast and acquisition scheme. |
| 2687 | Computer 51
|
Demonstration of an end-to-end open-source pipeline for a spiral-based rs-fMRI sequence |
| Marina Manso Jimeno1,2, John Thomas Vaughan Jr.1,2, and Sairam Geethanath2 | ||
1Department of Biomedical Engineering, Columbia University, New York, NY, United States, 2Columbia Magnetic Resonance Research Center, New York, NY, United States |
||
Spiral trajectories are suitable for fMRI as they provide high temporal resolution and motion tolerance. However, they are not typically available on scanners as the associated reconstruction is more involved. We have demonstrated the feasibility of an open-source spiral sequence end-to-end pipeline and compared it to the vendor-supplied EPI sequence. Although the spiral images showed signal loss near the frontal sinus, both datasets presented similar image quality and contrast. Importantly, the spiral sequence had a 2.3-fold advantage on temporal SNR and achieved analogous functional connectivity maps of the DMN, demonstrating the sequence BOLD sensitivity and viability.
|
| 2688 | Computer 52
|
Hybrid PCA denoising - improving PCA denoising in the presence of spatial correlations |
| Rafael Neto Henriques1, Sune Nørhøj Jespersen2,3, and Noam Shemesh1 | ||
1Champalimaud Research, Champalimaud Centre for the Unknown, Lisbon, Portugal, 2Center of Functionally Integrative Neuroscience (CFIN) and MINDLab, Clinical Institute, Aarhus University, Aarhus, Denmark, 3Department of Physics and Astronomy, Aarhus University, Aarhus, Denmark |
||
PCA denoising based on the Marchenko-Pastur (MP) distribution has become the state-of-the-art procedure to suppress thermal noise in multi-dimensional MRI. Here we developed a Hybrid-PCA strategy that combines a-priori noise variance estimation and the random matrix theory for PCA eigenvalue classification, to overcome shortcomings of contemporary MP-PCA denoising. Our results show that, while the MP-PCA denoising fails to classify the noise PCA components in data with spatially correlated noise, the Hybrid-PCA algorithm maintains its denoising performance. The Hybrid-PCA denoising can thus be a useful procedure for data corrupted by spatially correlated noise, as typically arises in vendor reconstructed data. |
| 2689 | Computer 53
|
seeVR: an open toolbox for analyzing cerebro-vascular reactivity (CVR) data |
| Alex A Bhogal1 | ||
1Radiology, UMC Utrecht, Utrecht, Netherlands |
||
The seeVR toolbox is a series of Matlab functions designed help analyze CVR through rapid implementation of processing pipelines. Depending on the type of stimulus used and the properties of the data acquisition, a number of options are available. All code is open-source and freely downloadable from the seeVR github repository (https://github.com/abhogal-lab/seeVR). |
| 2690 | Computer 54
|
BIDScoin: A user-friendly application to convert imaging data to the Brain Imaging Data Structure |
| Marcel Zwiers1, Cyril Pernet2, Anthony Galassi3, and Robert Oostenveld1,4 | ||
1Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, Netherlands, 2Neurobiology Research Unit, Copenhagen University Hospital, Copenhagen, Denmark, 3Center for Multimodal Neuroimaging, National Institute of Mental Health, Bethesda, MD, United States, 4NatMEG, Karolinska Institutet, Stockholm, Sweden |
||
Sharing neuroimaging data is an important development that has the potential to be scaled up with the new Brain Imaging Data Structure (BIDS) standard. Existing tools to converting data to BIDS often require programming skills or are tailored to specific institutes, datasets or data formats. Here we introduce BIDScoin, a cross-platform, flexible, free and open-source converter that provides a graphical user interface to help users finding their way in the BIDS standard, and supports plugins to extend its functionality. We show its design and demonstrate how it can be applied to a downloadable tutorial dataset. |
| 2691 | Computer 55
|
Open Science Initiative for Perfusion Imaging (OSIPI): A community-led, open-source code library for analysis of DCE/DSC-MRI |
| Petra J van Houdt1, Sudarshan Ragunathan2, Michael Berks3, Zaki Ahmed4, Lucy Kershaw5, Oliver Gurney-Champion6, Sirisha Tadimalla7, Jesper Kallehauge8, Jonathan Arvidsson9, Ben Dickie10, Simon Lévy11, Laura Bell12, Steven Sourbron13, and Michael Jonathan Thrippleton14 | ||
1Department of Radiation Oncology, the Netherlands Cancer Institute, Amsterdam, Netherlands, 2Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, 3Quantitative Biomedical Imaging Laboratory, Division of Cancer Sciences, University of Manchester, Manchester, United Kingdom, 4Radiology, Mayo Clinic, Rochester, MN, United States, 5Edinburgh Imaging and Centre for Cardiovascular Science, University of Edinburgh, Edinburgh, United Kingdom, 6Department of Radiology and Nuclear Medicine, Amsterdam UMC, Amsterdam, Netherlands, 7Institute of Medical Physics, The University of Sydney, Sydney, Australia, 8Department of Oncology, Aarhus University, Aarhus, Denmark, 9Institute of Clinical Sciences, Department of Radiation Sciences, The University of Gothenburg, Gothenburg, Sweden, 10Manchester Academic Health Science Centre, Division of Informatics, Imaging, and Data Science, School of Health Sciences, Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom, 11Institute of Radiology, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany, 12Clinical Imaging, Genentech, Inc, South San Francisco, CA, United States, 13Department of Infection, Immunity and Cardiovascular Disease, University of Sheffield, Sheffield, United Kingdom, 14Edinburgh Imaging and Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom |
||
A lack of validated, open-source code reduces the reliability of perfusion MRI, resulting in duplicate development. To address this problem, the Open Science Initiative for Perfusion Imaging (OSIPI) established a taskforce to collect, validate and harmonise such code. To date, 74 code contributions have been collected, with 14 of these tested. Source code and tests are published in an open-access repository. The OSIPI DCE/DSC-MRI code collection constitutes a valuable resource for researchers, and will ultimately be developed into a standardised, community-driven open-source code library. |
| 2692 | Computer 56
|
Visualising and interacting with diffusion MRI microstructural models using FSL-DIVE |
| Michiel Cottaar1, Nicole Eichert1, Hossein Rafipoor1, Amy Howard1, Daniel Kor1, Jiewon Kang1, Pavan Chaggar1,2, Ying-Qiu Zheng1, Sean Fitzgibbon1, and Saad Jbabdi1 | ||
1University of Oxford, OXFORD, United Kingdom, 2Mathematics, University of Oxford, Oxford, United Kingdom |
||
We introduce FSL-DIVE, an interactive visualisation software for diffusion microstructure modelling. This software allows the user to (i) flexibly set up a multi-compartment microstructure model, including fibre orientation modelling and restricted diffusion, (ii) choose different acquisition parameters, (iii) change the model parameters and analyse the model behaviour through multiple bespoke plots, (iv) save the predicted data (including noise) for any specified acquisition scheme. We hope this tool may be useful for teaching diffusion microstructure modelling, for developing new models, or for getting a better understanding of existing models. |
| 2693 | Computer 57
|
A model for fat-suppressed variable flip angle T1-mapping and dynamic contrast enhanced MRI |
| Myrte Wennen1,2, Manon Moll3, Tim Marcus2, Joost Kuijer2, Leo Heunks1, Christina Lavini2, Gustav Strijkers2, Aart Nederveen2, and Oliver Gurney-Champion2 | ||
1Intensive Care, Amsterdam UMC, Amsterdam, Netherlands, 2Radiology and Nuclear Medicine, Amsterdam UMC, Amsterdam, Netherlands, 3University of Twente, Enschede, Netherlands |
||
Although fat suppression can substantially improve image quality, the pulses corrupt the steady-state in a spoiled GRE sequence used for DCE and T1-relaxometry. Consequently, the conventional signal equations for quantifying tissue properties (T1, Ktrans, etc) do not hold. We have now included the effect of fat-suppression pulses in the signal equation that forms the basis for T1-relaxometry and DCE modelling. We have validated our equation with T1-mapping in a phantom and show its performance in healthy subjects and patients. Our equation enables accurate T1-mapping and pharmakinetic modelling for sequences that use fat suppression, including GRASP. |
| 2694 | Computer 58
|
Spin-Lock Times selection for Optimized T1ρ-Mapping of Knee Cartilage on Bi-Exponential and Stretched-Exponential Models |
| Hector Lise de Moura1, Marcelo Victor Wust Zibetti1, and Ravinder Regatte1 | ||
1Radiology, NYU Langone Health, New York, NY, United States |
||
T1ρ-Mapping requires acquisition with multiple Spin-Lock Times (TSLs) in order to fit the free parameters of the model. Reducing the #TSLs poses a trade-off between acquisition time and estimation error. Also, the choice of TSLs has a significant influence on the fitting results. A previous study analyzed the Cramér-Rao Lower Bound (CRLB) as an optimization criterium for the TSL choices using different #TSLs for Mono-Exponential models. This study extends the use of CRLB for Stretched- and Bi-exponential models using complex-valued fitting via Non-linear Least Squares. Experiments show that optimized TSL sequences improve parameter estimation when compared to non-optimized TSL sequences. |
| 2695 | Computer 59
|
Multivariate analysis of morphometric and quantitative magnetic resonance imaging metrics in aging and Alzheimer’s disease |
| Aurelie Bussy1,2, Raihaan Patel1,3, Alyssa Salaciak1, Sarah Farzin1, Stephanie Tullo1,2, Sandra Pelleieux4, Sylvia Villeneuve4,5, Judes Poirier4,5, John CS Breitner4,5, Gabriel A. Devenyi1,5, Christine L. Tardif3,6, and M. Mallar Chakravarty1,2,3,5 | ||
1Computational Brain Anatomy (CoBrA) Laboratory, Cerebral Imaging Centre, Douglas Mental Health University Institute, Montreal, QC, Canada, 2Integrated Program in Neuroscience, McGill University, Montreal, QC, Canada, 3Department of Biomedical Engineering, McGill University, Montreal, QC, Canada, 4Centre for the Studies on the Prevention of AD, Douglas Mental Health University Institute, Montreal, QC, Canada, 5Department of Psychiatry, McGill University, Montreal, QC, Canada, 6McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, QC, Canada |
||
Morphometric and quantitative magnetic resonance imaging techniques have rarely been used simultaneously to characterise healthy aging and Alzheimer’s disease (AD) progression. Here, we are extracting four vertex-wise cortical metrics : cortical thickness, surface area, T1 value (myelin) and T2* values (iron). All these metrics were analysed using non-negative matrix factorization and linear models. Overall, cortical thinning seemed to be linked to both aging and AD progression, while decrease in myelin seemed to be a phenomenon mostly related to aging. No significant patterns of changes were seen in the prodromal phase of AD. |
| 2696 | Computer 60
|
Characterising the impact of distortion correction in high-resolution 3D GRE R2* mapping at 7T. |
| Barbara Dymerska1, Oliver Josephs1, and Martina Callaghan1 | ||
1Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom |
||
R2*-mapping delivers quantitative information about tissue microstructure. We investigate the scale of image distortions in high-resolution, bipolar, multi-echo GRE acquisitions at 7T and their effects on R2*-mapping. We study reduction in variance by applying field map-based distortion correction while also considering the confounding effect of local smoothing. We show that for regions with signal displacement > 0.5 voxels the reduction in variance comes largely, not from smoothing, but from the appropriate repositioning of the tissue signal. Therefore, correcting distortions, also in sub-voxel regime, improves data consistency in bipolar GRE acquisitions, facilitating more robust R2*-mapping in high-resolution laminar studies for instance. |
| 2697 | Computer 61
|
Analysis on Echo Placement for Multi-echo Proton Resonance Frequency Shift Thermometry |
| Sven Nouwens1, Kemal Sumser2, Maarten Paulides2,3, Bram de Jager1, and Maurice Heemels1 | ||
1Mechanical Engineering, Control Systems Technology, Technical University Eindhoven, Eindhoven, Netherlands, 2Radiotherapy, Erasmus MC, Rotterdam, Netherlands, 3Electrical Engineering, Electromagnetics for Care & Cure, Technical University Eindhoven, Eindhoven, Netherlands |
||
Proton resonance frequency shift based MR thermometry is widely used to non-invasively monitor thermal therapies in vivo. Multi-echo gradient-echo sequences are being studied to increase the effective temperature to noise ratio, however, choosing the optimal echo-times is non-trivial. We developed an optimization framework based on measurement noise models to compute the optimal echo placement for schemes that either correct or do not correct for the conductivity bias. In a 4-echo phantom experiment, we demonstrated a reduction in the noise standard deviation of 22% when comparing optimal to equidistant echo placements. |
| 2698 | Computer 62
|
MAGiC in glioma: Are pre-contrast quantitative MRI parameters different in tumors with versus without enhancement? |
| Laura Nunez-Gonzalez1, Karin A. van Garderen1,2, Marion Smits1,2, Jaap Jaspers2, Alejandra Méndez Romero2, Dirk H. J. Poot1, and Juan A. Hernandez-Tamames1,3 | ||
1Erasmus MC, Rotterdam, Netherlands, 2Erasmus MC Cancer Institute, Rotterdam, Netherlands, 3TU Delft, Delft, Netherlands |
||
We analyze quantitative values in glioma obtained with MAGiC, which allows obtaining quantitative T1, T2 and PD maps in one single acquisition of less than 6 minutes for a whole brain. The maps were obtained before contrast-agent injection in 14 patients with glioma. We investigated the possibility of characterizing tumor regions based on quantitative maps acquired without contrast-agent. The results showed significant differences among tumoral tissue, tissue with T1w-enhancement, and normal white matter. Voxel-wise, this allowed to distinguish tumoral tissue but did not allow to accurately predicting T1w-enhancement. However, promising results were found predicting T1w-enhancement inside the tumor. |
| 2699 | Computer 63
|
Comparisons of Bayesian against non-Bayesian algorithms in intravoxel incoherent motion (IVIM) diffusion-weighted imaging in breast cancer |
| Sai Man Cheung1, Wing Shan Wu1, Nicholas Senn1, Ravi Sharma2, Trevor McGoldrick2, Tanja Gagliardi1,3, Ehab Husain4, Yazan Masannat5, and Jiabao He1 | ||
1Institute of Medical Sciences, University of Aberdeen, Aberdeen, United Kingdom, 2Oncology Department, Aberdeen Royal Infirmary, Aberdeen, United Kingdom, 3Radiology Department, Royal Marsden Hospital, London, United Kingdom, 4Pathology Department, Aberdeen Royal Infirmary, Aberdeen, United Kingdom, 5Breast Unit, Aberdeen Royal Infirmary, Aberdeen, United Kingdom |
||
Intravoxel incoherent motion (IVIM) model approximates the tissue perfusion as a form of pseudo-diffusion, extracted as fast diffusion component in diffusion weighted imaging (DWI). Despite the central role of tissue perfusion in the angiogenesis in cancer, the clinical application of IVIM is hampered due to the susceptibility to noise and tendency of overfitting bi-exponential decay function. Recent introduction of Bayesian algorithm significantly enhanced the robustness and accuracy of IVIM method, renewing its potential as a clinical tool. We therefore set out to examine current IVIM algorithms in the context of neoadjuvant chemotherapy on patients with breast cancer preceding the treatment. |
| 2700 | Computer 64
|
Performance Analysis of Undersampling Strategies for Multi-Shell Diffusion MRI |
| Elif Aygun1,2 and Emine Ulku Saritas1,2,3 | ||
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, 2National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara, Turkey, 3Neuroscience Graduate Program, Bilkent University, Ankara, Turkey |
||
Multi-shell dMRI metrics quantify information on micro-properties of neural tissue and can be used as markers for neurological diseases. The protocols used to acquire dMRI data often have prolonged acquisition times. In this work, we propose different undersampling strategies that reduce the acquisition time to half, and evaluate how the performances of multi-shell dMRI metrics change under these strategies. The results show that, while the best performing strategy changes for each metric, 3-shell gradient schemes with small variance of b-vector density on consecutive shells demonstrate improved performance. Additionally, more complex dMRI metrics exhibit relatively increased sensitivity to undersampling. |
| 2701 | Computer 65
|
T2 mapping for contiguous multi-slice 3D volume: A new EMC-based technique taking into account slice cross-talk effects |
| Ekaterina A Brui1, Charles A de Mayenne 2, Stanislas Rapacchi 3, Thomas Troalen 4, Christophe Vilmen3, and David Bendahan1,3 | ||
1School of Physics and Engineering, ITMO University, Saint-Petersburg, Russian Federation, 2National Higher French Institute of Aeronautics and Space, Toulouse, France, 3Aix-Marseille Universite, CNRS, Centre de Résonance Magnétique Biologique et Médicale, Marseille, France, 4Siemens Healthcare SAS, Saint-Denis, France |
||
Previously, an elegant method based on the generation of echo modulation curves (EMC) demonstrated a very high accuracy of T2 measurements. However, the original approach considered EMC dictionary modelling only for a single slice mode of multi-echo spin echo (MESE) sequence. For some applications, such as T2 mapping of cartilage in small joints, contiguous MRI is required. As a result, slice cross-talk effect may affect the accuracy of T2 estimation. In this work, we report a modified version of the EMC technique, which can be used to quantify T2 values for zero-gap multi-slice MESE acquisitions in tissues with long T1. |
| 2702 | Computer 66
|
Nanoscopic materials for quantitative water exchange phantoms |
| Scott D. Swanson1, Dariya I. Malyarenko1, and Thomas L. Chenevert1 | ||
1Radiology, University of Michigan, Ann Arbor, MI, United States |
||
Nanoparticle lipid vesicles are developed with controllable water exchange to provide a ground-truth benchmark for MR methods which measure exchange. Diffusion data are collected as a function of diffusion delay Δ. An exchange time, τ, is measured to be 164 ms in a phantom with a highly fluid membrane. These materials should help to clarify water dynamics in complex systems. |
| 2703 | Computer 67
|
Identifying perfused tissue regions through IVIM and ADC model comparison: technical validation |
| Damien J McHugh1,2, Michael J Dubec1,2, James Price2,3, Chris Moore1, David L Buckley1,4, James P. B. O'Connor2,5,6, David J Thomson3, and Andrew McPartlin3 | ||
1Christie Medical Physics and Engineering, The Christie NHS Foundation Trust, Manchester, United Kingdom, 2Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom, 3Department of Clinical Oncology, The Christie NHS Foundation Trust, Manchester, United Kingdom, 4Biomedical Imaging, University of Leeds, Leeds, United Kingdom, 5Department of Radiology, The Christie NHS Foundation Trust, Manchester, United Kingdom, 6Division of Radiotherapy and Imaging, The Institute of Cancer Research, London, United Kingdom |
||
Intravoxel incoherent motion (IVIM) can provide information about tissue perfusion in tumours and organs at risk, but the model may not be applicable in all voxels in a region of interest. Here, a model comparison framework is developed, based on comparing IVIM and apparent diffusion coefficient model fits, aiming to identify perfused voxels and quantify the proportion of tissue containing a perfusion component, pIVIM. Technical validation of the framework is performed, with simulations and phantom data showing that SNR and diffusion properties can bias pIVIM, while in vivo parotid gland data show that pIVIM exhibits good repeatability. |
| 2704 | Computer 68
|
Validation of semi-automated analysis of pancreas MRI-PDFF using a scan re-scan cohort from the Long COVID-19 study (COVERSCAN) |
| Alexandre Triay Bagur1, Michael Brady2, Arun Jandor2, Paul Aljabar2, and Daniel Bulte1 | ||
1Department of Engineering Science, University of Oxford, Oxford, United Kingdom, 2Perspectum Ltd, Oxford, United Kingdom |
||
Impairment of the pancreas has been shown in initial findings of the Long COVID-19 study (COVERSCAN), that uses quantitative MRI and expert manual analysis of the pancreas. Quantitative MRI analysis in such large-scale studies benefits from the decision support made possible by advanced image analysis methods. In this work, we validate a semi-automated pancreas processing pipeline that involves manual slice selection and automated segmentation and MRI-PDFF quantification. We show good agreement of the semi-automatic method with the reference manual processing by experts, as well as repeatable quantification in a scan re-scan subset of healthy subjects within COVERSCAN (n=35). |
| 2769 | Computer 37
|
TensorFlow MRI: A Library for Modern Computational MRI on Heterogenous Systems |
| Javier Montalt-Tordera1, Jennifer Steeden1, and Vivek Muthurangu1 | ||
1University College London, London, United Kingdom |
||
We present TensorFlow MRI, a new open-source library of TensorFlow operators for MR image reconstruction and processing. Its goal is to enable fast prototyping of modern MRI applications within a single computing framework. It is intended for researchers working in MR image reconstruction and/or processing, especially those interested in ML applications. The library is primarily Python-based and is easy to use, understand and extend. It has a single-command installation procedure and extensive documentation. Thanks to the use of a TensorFlow backend, it has excellent performance and runs on heterogenous and distributed systems. |
| 2770 | Computer 38
|
Constraint-Based Sequence Optimization in a Scanner-Independent MRI Framework |
| Daniel Christopher Hoinkiss1, Simon Konstandin1,2, and Matthias Günther1,2,3 | ||
1Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany, 2mediri GmbH, Heidelberg, Germany, 3University of Bremen, Bremen, Germany |
||
We demonstrate a constraint-based approach of MRI sequence development in the vendor-independent MRI pulse sequence development framework gammaSTAR and demonstrate this concept in arterial spin labeling by optimizing the timings of the background suppression pulses to minimize the signal contribution of chosen T1 values in the human brain. This concept can raise MRI sequence development to a new abstraction level by, instead of providing exact timings and parameter values, defining physical constraints and conditions to be satisfied by the MRI sequence during an automated sequence generation process. |
| 2771 | Computer 39
|
Efficient Non-Uniform Fourier Transform with embedded low-rank projection for fast model-based MRI reconstruction |
| Matteo Cencini1,2, Luca Peretti2,3, and Michela Tosetti1,2 | ||
1IRCCS Stella Maris, Pisa, Italy, 2Imago7 Foundation, Pisa, Italy, 3Università di Pisa, Pisa, Italy |
||
Modern approaches to iterative imaging, such as model-based reconstruction, requires efficient implementations of Non-Uniform Fourier Transform to reach feasible reconstruction times. In addition, low-rank subspace projection is often used to reduce computational burden. While many implementations of NUFFT currently exists, they are not optimized for this kind of problems. Here, we propose a fast and memory-efficient NUFFT operator with embedded low-rank subspace projection. We demonstrate an order of magnitude of speed-up with comparable image quality compared to other high-level implementations. |
| 2772 | Computer 40
|
Flow 2.0 - a flexible, scalable, cross-platform post-processing software for real-time phase contrast sequences |
| Liu Pan1,2, Sidy Fall1, and Olivier Baledent1,2 | ||
1CHIMERE UR 7516, Jules Verne University, Amiens, France, 2Medical Image Processing Department, Jules Verne University Hospital, Amiens, France |
||
Real-time phase contrast sequences (RT-PC) has potential value as a scientific and clinical tool in quantifying the effects of respiration on cerebral circulation. To simplify its complicated post-processing process, we developed Flow 2.0 software, which provides a complete post-processing workflow including converting DICOM data, image segmentation, image processing, data extraction, background field correction, anti-aliasing filter, signal processing and analysis and a novel time-domain method for quantifying the effect of respiration on the cerebral circulation. This end-to-end software allows us to quickly, robustly and accurately perform batch process RT-PC and multivariate analysis of the effects of respiration on cerebral circulation. |
| 2773 | Computer 41
|
A Cloud-based MR Spectroscopy Tool for Basis Set Simulation |
| Steve C.N. Hui1,2, Muhammad G. Saleh3, Helge J. Zöllner1,2, Georg Oeltzschner1,2, Hongli Fan1,4, Yue Li5, Yulu Song1,2, Hangyi Jiang1, Jamie Near6,7, Susumu Mori1,2, Hanzhang Lu1,2, and Richard A.E. Edden1,2 | ||
1Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States, 2F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, 3Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD, United States, 4Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, United States, 5AnatomyWorks, LLC, Ellicott City, MD, United States, 6Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada, 7Sunnybrook Research Institute, Toronto, ON, Canada |
||
MRSCloud is an online-based spectral simulation tool for brain metabolites. Up to 32 metabolites can be simulated to generate a basis set for linear combination modeling. The 1D projection method, coherence pathway filters and pre-calculation of propagators have been implemented for density-matrix simulations to run on a high-performance cloud server. Results indicated simulations were comparable with FID-A and difference between simulations of spatial points (21x21/41x41/101x101) were small. It allows community users to generate vendor, sequence, editing experimental-specific basis sets that are appropriate for their studies. |
| 2774 | Computer 42
|
Mechanical Resonance Recording and Analysis using auDIO files (MR-RADIO) |
| Hannes Dillinger1, Eva Peper1, Christian Guenthner1, and Sebastian Kozerke1 | ||
1ETH and University Zurich, Zurich, Switzerland |
||
This work presents a web-based framework for the identification of mechanical resonances of MRI gradients systems based exclusively on mobile phone audio recordings. frequencies are used to guide beneficial combinations of TE/TR for PC-MRI and readout bandwidth settings for EPI sequences. Results demonstrate that the background phase in PC-MRI as well as beat phenomena during EPI readouts can be reduced. The web app can be found at https://mr-radio.herokuapp.com and its source code is publicly available at https://github.com/hdillinger/mr-radio in the spirit of reproducible research. |
| 2775 | Computer 43
|
Virtual Scanner 2.0: enabling the MR digital twin |
| Gehua Tong1, John Thomas Vaughan, Jr. 2, and Sairam Geethanath2 | ||
1Biomedical Engineering, Columbia University, New York, NY, United States, 2Columbia Magnetic Resonance Research Center, Columbia University, New York, NY, United States |
||
Virtual Scanner is an open-source MR platform for end-to-end simulation and serves as a digital twin for real MR scanners in hardware and software aspects. In this update, we present new functions including RF pulse design and simulation, hardware-incorporated Bloch simulation that takes into account B0 and B1- fields, and non-Cartesian reconstruction. Three Google Colab notebooks are provided to demonstrate these functions and enable users to perform virtual experiments in a zero-footprint manner. Furthermore, we demonstrate the digital twin concept in a development iteration case using the SPGR sequence. |
| 2776 | Computer 44
|
ALFONSO: A versatiLe Formulation fOr N-dimensional Signal mOdel fitting of MR spectroscopy data and its application in MRS of body lipids |
| Stefan Ruschke1 and Dimitrios C. Karampinos1,2 | ||
1Department of Diagnostic and Interventional Radiology, Technical University of Munich, Munich, Germany, 2Munich Institute of Biomedical Engineering, Technical Universit of Munich, Munich, Germany |
||
Multi-dimensional MR spectroscopy is frequently used for the probing of MR properties including the characterization of the water–fat environment in body applications. Previously, many studies used an ad-hoc multi-step approach where the fitting of the spectral content and the signal modelling was performed in separate steps albeit the theoretically compromised precision when compared to a joint fitting and modelling approach. ALFONSO allows the convenient and yet flexible definition of joint fitting and modelling strategies. Its utility and supremacy are demonstrated for the quantification of fat fraction in the liver, ADC in bone marrow and lipid droplet size in a phantom |
| 2777 | Computer 45
|
An open-source repository of minimum echo time sequences |
| Gehua Tong1, John Thomas Vaughan, Jr. 2, and Sairam Geethanath2 | ||
1Biomedical Engineering, Columbia University, New York, NY, United States, 2Columbia Magnetic Resonance Research Center, Columbia University, New York, NY, United States |
||
Ultrashort Echo Time (UTE) sequences are clinically useful in short T2 tissues such as bone, cartilage, and lung parenchyma. To help harmonize UTE protocols across clinical sites, we present an open-source repository implementing three basic radial minimal TE sequences using PyPulseq: 2D UTE, 3D UTE, and COncurrent Dephasing and Excitation (CODE). Functions for customized sequences and reconstruction scripts were provided for a complete pipeline. Echo Times (TEs) from 0.12 to 0.82 ms were achieved, and 27.6% more signal was captured in the bone and connective tissue in 2D ex vivo images compared to the GRE reference. |
| 2778 | Computer 46
|
Ginkgo: a novel modular and and Open Source MRI pulse sequence development framework dedicated to MRI systems |
| Anaïs Artiges1, Franck Mauconduit1, Ivy Uszynski1, Baptiste Mulot2, Elodie Chaillou3, Philippe Ciuciu4, and Cyril Poupon1 | ||
1BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA, Gif-sur-Yvette, France, 2Beauval Nature, ZooParc de Beauval, Saint-Aignan, France, 3UMR 85 Physiologie de la Reproduction et des Comportements, INRAE Centre Val-de-Loire, Nouzilly, France, 4Parietal, NeuroSpin, Paris-Saclay University, Inria, CEA, Gif-sur-Yvette, France |
||
We created a new object-oriented environment for MR pulse-sequence development based on IDEA VE11C and above versions using an Open Science philosophy. This Ginkgo toolkit uses a modular structure to facilitate the design of pulse sequences using the aggregation of basic open-source sequence blocks available from the toolkit. Proofs of concept of the productivity gain reached using Ginkgo are provided through the implementation of a series of sequence models including a diffusion-weighted PGSE 3D EPI sequence. |
| 2779 | Computer 47
|
Realistic Simulation of Body Composition MRI and Field-Strength Dependence |
| Ecrin Yagiz1, Sophia X. Cui2, Nam G. Lee1, Bilal Tasdelen1, Ye Tian1, and Krishna S. Nayak1,3 | ||
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, 2Siemens Medical Solutions USA, Inc.,, Los Angeles, CA, United States, 3Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, United States |
||
MRI is used extensively for body composition assessment in large part due to the increasing prevalence of metabolic disease and obesity. In this work, we provide a framework for simulating body composition MRI, including non-Cartesian sequences and emerging low field strengths. This framework can be utilized to investigate the optimality of acquisition parameters (e.g., resolution, TE, sampling trajectory) and the performance of fat/water separated reconstruction algorithms. |
| 2780 | Computer 48
|
A software tool to assess radiofrequency coil designs with respect to ultimate intrinsic performance limits |
| Eros Montin1, Ioannis P Georgakis1,2, Bei Zhang3, and Riccardo Lattanzi1,4 | ||
1Center for Advanced Imaging Innovation and Research (CAI2R) Department of Radiology, New York University Grossman School of Medicine, New York, NY, United States, 2Corsmed, Stockolm, Sweden, 3Advanced Imaging Research Center, UT Southwestern Medical Center, Dallas, TX, United States, 4Vilcek Institute of Graduate Biomedical Sciences, New York University Grossman School of Medicine,, New York, NY, United States |
||
This work introduces POIROT (Performance Observer In Receive or Transmit), a web-based tool for the assessment of receive and transmit coil designs against ultimate intrinsic signal-to-noise ratio and transmit efficiency, respectively. It enables engineers, for the first time, to evaluate how good is a design and whether there is further room for improvement before building a prototype. POIROT currently includes ultimate intrinsic data for three numerical head models at different resolutions and magnetic field strengths. POIROT could be integrated with rapid numerical EM modeling tools to develop a pipeline for coil design optimization that uses ultimate performance as the benchmark. |
| 2781 | Computer 49
|
A BIDS extension proposal for magnetic resonance spectroscopy data |
| Mark Mikkelsen1, Dickson Wong2, Wolfgang Bogner3, Yaroslav O. Halchenko4, Damon G. Lamb5,6,7,8,9, Paul G. Mullins10, Georg Oeltzschner11,12, and Martin Wilson13 | ||
1Department of Radiology, Weill Cornell Medicine, New York, NY, United States, 2Schulich School of Medicine and Dentistry, Western University, London, ON, Canada, 3High-field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 4Department of Psychological and Brain Sciences, Dartmouth College, Hanover, NH, United States, 5Departments of Psychiatry, Neuroscience, and Biomedical Engineering, University of Florida, Gainesville, FL, United States, 6Center for OCD, Anxiety, and Related Disorders, University of Florida, Gainesville, FL, United States, 7Center for Cognitive Aging and Memory, University of Florida, Gainesville, FL, United States, 8McKnight Brain Institute, University of Florida, Gainesville, FL, United States, 9Brain Rehabilitation Research Center, Malcom Randall Veterans Affairs Medical Center, Gainesville, FL, United States, 10Bangor Imaging Unit, School of Psychology, Bangor University, Bangor, United Kingdom, 11Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, 12F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, 13Centre for Human Brain Health and School of Psychology, University of Birmingham, Birmingham, United Kingdom |
||
Magnetic resonance spectroscopy (MRS) studies can generate extensive and complex datasets. Datasets can be organized idiosyncratically across (and even within) labs. However, sharing data and reproducing results without the added challenge of parsing lab-specific file organization is paramount to open science. Sharing of MRS data that is independent of lab-specific practices would greatly aid transparency and reusability. Here, we present an extension to the Brain Imaging Data Structure (BIDS) specification for MRS data. MRS-BIDS adopts the established tenets of standardization of BIDS and applies them to MRS data while considering the nuances relevant to MRS methodology. |
| 2782 | Computer 50
|
Study of potential immediate effects of neuronal activity on the MRI signal in humans |
| Peter van Gelderen1, Jacco A de Zwart1, and Jeff H Duyn1 | ||
1Advanced MRI section, LFMI, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, United States |
||
A recent study has shown rapid MRI signals in response to whisker pad stimulation in mice that track electrical activity. We used rapid, high SNR imaging at 7 T to look for a possible direct effect of visual stimulation on MRI signal in human visual cortex. Two rapid stimulation protocols and variation of T1-weighting failed to generate observable signal changes above the 0.01% detection threshold, well below the previously reported effect size of 0.15%. This confirms the well-established difficulty in developing more direct measures of neuronal activity than available with BOLD-fMRI. |
| 2783 | Computer 51
|
A Software Tool for Inline Post-processing and Visualization of Chemical Exchange Saturation Transfer (PV-CEST) on 3T Scanner |
| Chuyu Liu1, Yishi Wang2, Zhensen Chen3, Yajing Zhang4, and Xiaolei Song1 | ||
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, Tsinghua University, Beijing, China, 2Philips Healthcare, Beijing, China, 3Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China, 4MR Clinical Science, Philips Healthcare, Suzhou, China |
||
We developed a real-time software for CEST post-processing and visualization on Philips 3T scanner, termed as PV-CEST. PV-CEST runs in the vendor provided PRIDE 2.0 environment, providing image view, Z-spectra analysis and quantitative maps calculation functionalities during scanning procedure, which could have great clinical utility. |
| 2784 | Computer 52
|
PySynthMRI: An open-source Python tool for Synthetic MRI |
| Luca Peretti1,2,3, Matteo Cencini3, Paolo Cecchi2,4, Graziella Donatelli2,4, Mauro Costagli3,5, and Michela Tosetti3 | ||
1University of Pisa, Pisa, Italy, 2Imago7 Research Foundation, Pisa, Italy, 3IRCCS Stella Maris, Pisa, Italy, 4Azienda Ospedaliero-Universitaria Pisana, Pisa, Italy, 5University of Genoa, Genoa, Italy |
||
Synthetic MR imaging allows reconstruction of different image contrasts from a single acquisition, reducing scan times. Here we introduce PySynthMRI, an open-source tool which provides a user-friendly and fluid interface to synthesize and modify contrast-weighted images. The tool allows the user to add and adjust virtual scanner parameters showing, in real time, how the resulting synthesized images vary accordingly, which makes PySynthMRI a valuable tool for both research and teaching. The generated images can be exported in the preferred format. |
| 2785 | Computer 53
|
Multi-center, Multi-vendor evaluation of quantitative susceptibility mapping for quantification of liver iron |
| Collin J. Buelo1,2, Ruiyang Zhao1,2, Julia V. Velikina1, Steffen Bollmann3, Ante Zhu4, Qing Yuan5, Mounes Aliyari Ghasabeh6, Stefan Ruschke7, Dimitrios C. Karampinos7, David T. Harris1, Ryan J. Mattison8, Michael R. Jeng9, Ivan Pedrosa5, Ihab R. Kamel5, Shreyas Vasanawala10, Takeshi Yokoo5,11, Scott B. Reeder1,2,8,12,13, and Diego Hernando1,2 | ||
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, 2Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, 33School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane, Australia, 4GE Global Research, Niskayuna, NY, United States, 5Radiology, University of Texas Southwestern Medical Center, Dallas, TX, United States, 6Radiology, The Johns Hopkins University, Baltimore, MD, United States, 7Department of Diagnostic and Interventional Radiology, Technical University of Munich, Munich, Germany, 8Medicine, University of Wisconsin-Madison, Madison, WI, United States, 9Pediatrics - Hematology & Oncology, Stanford University, Palo Alto, CA, United States, 10Radiology, Stanford University, Palo Alto, CA, United States, 11Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, TX, United States, 12Department of Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States, 13Emergency Medicine, University of Wisconsin-Madison, Madison, WI, United States |
||
Quantitative susceptibility mapping (QSM) of the liver is an emerging method to quantify liver iron concentration (LIC). Multiple new methods have recently been introduced for QSM in the liver. These methods include CobraChi, a deep learning-based method, as well as an L2-optimization based method. This work evaluates the performance of these liver QSM methods across multiple centers and MRI vendors. |
| 2786 | Computer 54
|
Non-steady-state sequences for multi-parametric MRI need to be evaluated in the context of gradient-encoding |
| Miha Fuderer1,2, Oscar van der Heide1, Hongyan Liu1, Cornelis A.T. van den Berg1, and Alessandro Sbrizzi1 | ||
1Computational Imaging Group for MR Diagnostics and Therapy, Center for Image Sciences, University Medical Center Utrecht, Utrecht, Netherlands, 2Department of Radiology, Division of Imaging and Oncology, University Medical Center Utrecht, Utrecht, Netherlands |
||
In non-steady-state multi-parametric quantitative MRI (e.g. MR-STAT or MR Fingerprinting (MRF)), it is a non-trivial task to devise good sequences of time-varying flip angles. Recent work addresses this issue, albeit assuming a single-voxel approach, i.e. the k-space encoding is not taken into account when optimizing. In this work we show, using examples from MR-STAT reconstructions and using a Cramer-Rao based BLAKJac analysis, that two apparently similar sequences can have vastly different outcomes when applied in an actual (2-dimensional) measurement setup - showing that the context of encoding is very relevant. |
| 2787 | Computer 55
|
APT-CEST Scan-Rescan Reproducibility in Healthy Volunteers and Brain Glioma Patients at 3 Tesla |
| Ivar J.H.G. Wamelink1, Beatriz Padrela1, Joost P.A. Kuijer1, Yi Zhang2, Frederik Barkhof1,3, Henk J.M.M. Mutsaerts1, Elsmarieke van de Giessen1, and Vera C. Keil1 | ||
1Department of Radiology and Nuclear Medicine, Vrije Universiteit Amsterdam Medisch Centrum, Amsterdam, Netherlands, 2Department of Biomedical Engineering, Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, Zhejiang, China, 3UCL Institutes of Neurology and Healthcare Engineering, London, United Kingdom |
||
Amide proton transfer (APT) chemical exchange saturation transfer (CEST) MRI is a potentially useful clinical technique to image brain tumors, but its reproducibility was not yet investigated in detail. Our 3T scan-rescan protocol on volunteers and glioma patients revealed high within-session and slightly lower in-between-session and between-day reproducibility, with highest reproducibility occipitally and centrally in the brain and lowest at the skull base. The within-session reproducibility for patients was also good, both intratumoral and extratumoral. These results show that APT-CEST at 3T MRI renders reproducible values allowing for clinical monitoring of metabolic brain tumors. |
| 2788 | Computer 56
|
Repeatability of ultra-high-resolution Multi-Parametric Mapping across five 7T sites |
| siya sherif1, Ali Aghaeifar2,3, Kerrin Pine4, Belinda Ding5, Maryam Seif6, Evelyne Balteau1, Daniel Nanz7,8, Christine Bastin1, Eric Salmon1,9, Pierre Maquet1,9, Gilles Vandewalle1, Christopher T Rodgers 5, Patrick Freund6, Nikolaus Weiskopf4,10, Christophe Phillips1,11, and Martina F Callaghan3 | ||
1GIGA-Cyclotron Research Centre - In Vivo Imaging, University of Liège, Liège, Belgium, 2MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom, 3Wellcome Centre for Human Neuroimaging, University College London, London, United Kingdom, 4Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 5Wolfson Brain Imaging Centre, University of Cambridge, Cambridge, United Kingdom, 6Spinal cord injury Center, University of Zurich, Zurich, Switzerland, 7Swiss Center for Musculoskeletal Imaging, Zurich, Switzerland, 8University of Zurich, Zurich, Switzerland, 9Department of Neurology, University Hospital of Liège, Liège, Belgium, 10Felix Bloch Institute for Solid State Physics, Leipzig University, Leipzig, Germany, 11GIGA - In Silico Medicine, University of Liège, Liège, Belgium |
||
Repeatability is key to the utility of quantitative MRI, which promises standardised measures with biological relevance. Here we tested a candidate ultra-high resolution (0.6mm isotropic) multi-parameter mapping protocol at five 7T sites. Repeatability was good, with B0 and B1+ field inhomogeneities being the limiting factor. MTsat had the lowest repeatability and PD the highest. R1 and R2* were intermediate. Repeatability was also lowest in temporal and occipital cortices. These observations were consistent across sites. |
| 2789 | Computer 57
|
Test-retest repeatability of radiomic features derived from T2w MRI in prostate cancer patients |
| Barbara Daria Wichtmann1, Steffen Albert2, Daniel Pinto dos Santos3, Ulrike Irmgard Attenberger1, and Bettina Baessler4 | ||
1Department of Diagnostic and Interventional Radiology, University Hospital Bonn, Bonn, Germany, 2Computer Assisted Clinical Medicine, Mannheim Institute for Intelligent Systems in Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, 3Department of Radiology, University Hospital Cologne, Cologne, Germany, 4Institute of Diagnostic and Interventional Radiology, University Hospital Wuerzburg, Wuerzburg, Germany |
||
Radiomics enables the extraction of quantitative features from medical images, potentially augmenting the characterization of healthy and diseased tissue. Before these features can be routinely used as biomarkers in clinical practice, however, their repeatability and reproducibility must be ensured. This study seeks to investigate feature repeatability in an in-vivo, clinical test-retest dataset of prostate cancer patients. Our results show that the majority (71.8%) of radiomic features extracted from in-vivo, clinical T2-weighted images was not repeatable, emphasizing the need for repeatability and reproducibility studies. |
| 2790 | Computer 58
|
Reproducibility of rapid quantitative multiparameter mapping using skipped-CAIPI 3D-EPI |
| Difei Wang1, Rüdiger Stirnberg1, and Tony Stöcker1,2 | ||
1German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, 2Department of Physics and Astronomy, University of Bonn, Bonn, Germany |
||
We investigated the reproducibility of multiparameter mapping (MPM) using multi-echo skipped-CAIPI 3D-EPI at 3T. Compared to >15 minutes of FLASH-MPM, EPI-MPM under 3 minutes achieved considerably high repeatability. In only one fifth of the total scan time of FLASH, the resulting CoVs of EPI were only 1.2-2.2 times larger than FLASH for R1, PD and MTsat, while the CoV of EPI for R2* was even smaller. Minor differences of the observed parameter estimates can be attributed to the intrinsic difference between EPI and FLASH sequence timing. |
| 2791 | Computer 59
|
Quantitative Assessment of Diffusion Weighted Imaging Near Metal Implants |
| John P. Neri1, Matthew F. Koff1, Kevin M. Koch2, and Ek T. Tan1 | ||
1Hospital for Special Surgery, New York, NY, United States, 2Medical College of Wisconsin, Milwaukee, WI, United States |
||
Conventional echo-planar imaging (EPI) DWI suffers from substantial artifact when imaging orthopedic hardware. DWI with multi-acquisition with variable resonance image combination (MAVRIC) can reduce susceptibility artifact around metal implants. The purpose of this work was to evaluate the quantitative accuracy of DWI-MAVRIC near metal components using a diffusion phantom. DWI-MAVRIC sequences were acquired in the presence of no metal, cobalt-chromium, titanium, and stainless steel. It was found that DWI-MAVRIC can accurately measure apparent diffusion coefficient (ADC) in the presence of metal. Among tested metals, stainless steel creates the greatest artifact that prevented the acquisition of accurate ADC data. |
| 2792 | Computer 60
|
Heterogeneity of ASL perfusion MRI in low-grade paediatric glioma as imaging biomarker to assess treatment effect |
| Lejla Alic1, Sanneke C Willekens1,2, Henk-Jan M.M. Mutsaerts2, Jan Petr3, Netteke A.Y.N. Schouten-van Meeteren4,5, Maarten M.H. Lequin6, and Evita E.C. Wiegers2 | ||
1Magnetic Detection & Imaging Group, University of Twente, Enschede, Netherlands, 2Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands, 3Helmholtz-Zentrum Dresden-Rossendorf, Institute for Radiopharmaceutical Cancer Research, Dresden, Germany, 4Radiology and Nuclear Medicine, Amsterdam Neuroscience, Amsterdam, Netherlands, 5Department of Neuro-oncology, Princess Máxima Centre, Utrecht, Netherlands, 6Radiology, University Medical Center Utrecht, Utrecht, Netherlands |
||
ASL-MRI is reported as an option to assess potentially heterogeneous physiological processes important for tumour treatment. Therefore, we explored the heterogeneity in normalised CBF as an imaging biomarker for assessment of treatment effect in pLGG. There is a noticeable effect of chemotherapy observed as a change in texture of healthy appearing brain tissue. A high difference in texture between treated and non-treated patients for non-enhancing tumour part is observed, suggesting that texture, based on co-occurrence matrices, is suitable as an imaging biomarker for assessment of treatment effect in pLGG. |
| 2793 | Computer 61
|
Reduced field-of-view multi-shell DWI of the sciatic nerve: A reproducibility assessment |
| Ratthaporn Boonsuth1, Rebecca S Samson1, Francesco Grussu1,2, Marco Battiston1, Torben Schneider3, Masami Yoneyama4, Ferran Prados1,5,6, Carmen Tur1,7, Sara Collorone1, Rosa Cortese1, Claudia AM Gandini Wheeler-Kingshott1,8,9, and Marios C Yiannakas1 | ||
1NMR Research Unit, Queen Square MS Centre, Department of Neuroinflammation, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 2Radiomics Group, Vall d’Hebron Institute of Oncology, Vall d’Hebron Barcelona Hospital Campus, Barcelona, Spain, 3Philips Healthcare, Guildford, Surrey, United Kingdom, 4Philips Japan, Minatoku, Tokyo, Japan, 5Centre for Medical Image Computing, Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom, 6E-Heath Centre, Universitat Oberta de Catalunya, Barcelona, Spain, 7Multiple Sclerosis Centre of Catalonia (Cemcat), Vall d’Hebron Institute of Research, Vall d'Hebron Barcelona Hospital Campus, Barcelona, Spain, 8Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy, 9Brain Connectivity Research Centre, IRCCS Mondino Foundation, Pavia, Italy |
||
Advanced diffusion-weighted imaging (DWI) methods are powerful diagnostic and research tools when applied in the central nervous system (CNS). However, the use of DWI methods to study individual nerves outside the CNS in vivo, such as the sciatic nerve, has been hampered by a number of technical challenges. In this study, we explore the feasibility of acquiring multi-shell DWI metrics in the sciatic nerve using reduced field-of-view echo planar imaging and report results from a reproducibility assessment in in healthy controls. |
| 2794 | Computer 62
|
Psychophysical evaluation of radiologic vs. deep-learning based identification of multiple sclerosis brain lesions |
| Chen Solomon1, Omer Shmueli1, Tamar Blumenfeld-Katzir1, Dvir Radunsky1, Noam Omer1, Neta Stern1, Shai Shrot2,3, Moti Salti4,5, Hayit Greenspan1, and Noam Ben-Eliezer6,7 | ||
1Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel, 2Department of Diagnostic Imaging, Sheba Medical Center, Ramat Gan, Israel, 3Tel Aviv University, Tel Aviv, Israel, 4Brain Imaging Research Center, Soroka Medical Center, Beer Sheva, Israel, 5University Medical Center, Ben Gurion University, Beer Sheva, Israel, 6Center for Advanced Imaging Innovation and Research (CAI2R), New-York University Langone Medical Center, New York, NY, United States, 7Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel |
||
Computer assisted detection (CAD) of pathology in MRI scans may provide higher sensitivity to tissue changes. We present rigorous comparison of CAD vs. conventional radiologic evaluation of multiple sclerosis (MS) lesions. A psychophysical experiment was performed, where radiologists and a deep neural-network were asked to detect artificial MS lesions, synthetically simulated on T2-weighted FLAIR images, and at 8 levels of severity. Odds ratio analysis indicated that the human vision is less sensitive to low-severity lesions. This suggests that CAD can improve early detection of tissue abnormalities in the brain. |
| 2795 | Computer 63
|
Test-retest and inter-vendor variability of brain segmentation using 3D synthetic MRI in volunteers at 3T |
| Maarten Naeyaert1, Tim Vanderhasselt1, and Hubert Raeymaekers1 | ||
1Radiology, Universitair Ziekenhuis Brussel, Brussels, Belgium |
||
Six volunteers were scanned two times on two different 3T scanners (GE, Philips) using the 3D-QALAS sequence. The intracranial and brain parenchymal volume and myelin content were determined and the brain segmented in cerebrospinal fluid and white and grey matter using synthetic MRI. Bland-Altman plots, Dice coefficient and coefficient of variation indicate that the test-retest variability is very small, but there is an inter-vendor bias for CSF, GM and WM. MyC, BPV and ICV have only a very limited inter-vendor bias. At least for GM and BPV, the expected inter-scanner variation, besides the bias, is below the clinically relevant threshold. |
| 2796 | Computer 64
|
Repeatability and accuracy of MR-STAT quantitative T1 and T2 measurements |
| Oscar van der Heide1,2, Miha Fuderer1,2, Hongyan Liu1,2, Cornelis A.T. van den Berg1,2, and Alessandro Sbrizzi1,2 | ||
1Computational Imaging Group for MR Diagnostics and Therapy, Center for Image Sciences, University Medical Center Utrecht, Utrecht, Netherlands, 2Department of Radiology, Division of Imaging and Oncology, University Medical Center Utrecht, Utrecht, Netherlands |
||
MR-STAT is a quantitative MRI technique that obtains multiple tissue parameter maps (e.g. T1, T2) from a single, short scan. In this study we perform an accuracy and repeatability study of MR-STAT using a Cartesian FISP-based sequence by repeatedly scanning Eurospin gel tubes and comparing the results against reference measurements. We observe coefficients of variation between 0.5 % and 1.5 % for T1 and between 2.0 % and 7.2 % for T2. The mean relative errors as compared to the reference measurements are -1.5 % for T1 and 5.8 % for T2. |
| 2797 | Computer 65
|
Quantitative & qualitative evaluation of accelerated T2 mapping technique using Deep Learning reconstruction in knee cartilage |
| Laura Carretero1,2, Maggie Fung3, Pablo García-Polo1, Daniel Litwiller4, Marc Lebel5, Graeme C McKinnon6, and Mario Padrón7 | ||
1GE Healthcare, Madrid, Spain, 2Rey Juan Carlos University, Madrid, Spain, 3GE Healthcare, New York, NY, United States, 4GE Healthcare, Colorado, CO, United States, 5GE Healthcare, Calgary, AB, Canada, 6GE Healthcare, Waukesha, WI, United States, 7Clínica Cemtro, Madrid, Spain |
||
MRI-based cartilage imaging shows biochemical and microstructural changes at early stages of osteoarthritis before changes become visible with structural MRI and arthroscopy. However, clinical application of T2 mapping is driven by high variability and suboptimal reproducibility, besides long examination times. The purpose of this study is to accelerate T2 mapping technique and apply a novel DL reconstruction method to develop a fast and robust way to access T2 relaxometry. Phantom and in-vivo testing demonstrated that DL reconstructed accelerated images provide increased consistency compared to conventional reconstruction and implies a great step into an extensive clinical adoption. |
| 2798 | Computer 66
|
Within‐subject reliability of proton density estimation using quantitative magnetic resonance imaging |
| Karthik R Sreenivasan1, Dietmar Cordes1, Virendra Mishra1, and Le H Hua1 | ||
1Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, NV, United States |
||
New advances in quantitative MRI (qMRI) techniques provide basic MRI properties such as proton density (PD) that can act as a direct surrogate of the tissue integrity in the voxel. PD may help demonstrate differences on an individual level in patients with multiple sclerosis. In this study, we wanted to compare the within-subject reproducibility of qMRI-based PD measures. Variability in the estimated PD was low and fell within the limits for reliability. These results suggest that in-vivo estimates of PD using qMRI are reproducible within the subject in the same scanner and could play a vital role in clinical studies. |