Three-dimensional MRI is a valuable tool for the diagnosis of heart diseases, which furthermore offers advantages such as intrinsically higher signal-to-noise ratios, less scan planning efforts and the possibility to reconstruct arbitrary anatomical planes from one dataset. Despite these advantages, inherently long acquisition durations often limit clinical applications. This abstract highlights the possibility of acquiring qualitative 3D images at an isotropic resolution of 1.09 mm within 250 heartbeats (ECG-gating), using specifically generated k-space interleaves with low-coherent and low-discrepancy sampling properties, based on a generalised form of the previously introduced Seiffert-Spirals. Cardiac image segmentations are furthermore presented to emphasise image qualities.

Calcifications in atherosclerotic plaques are possibly associated with worsened clinical prognosis. In contrast to the often-used CT for diagnosis, calcifications cannot easily be detected with MRI. As PET becomes a major complementary modality for MRI in plaque imaging, and as integrated PET/MR scanners become increasingly available, PET/MR based calcification screening methods are desired. With zero TE (ZTE) MR imaging it is possible to image short T2 tissues like bone. In the current study we demonstrate that ZTE imaging on PET/MR, combined with Deep Learning reconstructed ZTE, can likely be applied to detect calcifications in peripheral vasculature.

We propose Bayesian accelerated Compressed Sensing (BaCS) to improve the computational speed of Bayesian CS by two orders of magnitude. We achieve this by circumventing a costly matrix inversion problem using conjugate gradients and Monte Carlo sampling, which lend themselves well to parallel processing using GPUs. Exploiting parallelism renders BaCS even faster than sparseMRI, while having the ability to exploit similarities between multi-contrast images to improve reconstruction performance. Further, we extend BaCS to multi-channel reconstruction by synergistically combining it with SENSE to enable yet higher acceleration rates. 

Parallel Imaging (PI) is a crucial technique for accelerating data acquisition in Magnetic Resonance Imaging (MRI), which is exceedingly time-consuming. With current SENSE-based MRI reconstruction formulated as a trainable unrolled optimization framework with several cascades of regularization networks and varying data consistency layers, coils sensitivity maps (CSMs) are needed at each cascade. Therefore, we propose a deep sets CSM estimation network (DS-CSME in short), enabling an end-to-end deep learning solution that allows for further MRI acceleration while preserving the overall reconstructed image quality.

FID sequences such as the zero echo time sequence (ZTE) posses many advantages including capturing signals from fast relaxing spins, efficient use of time and quite acoustic operation. They possess one major disadvantage, information from the start of the FID is nearly always missing or corrupted due to requirements for RF pulse time and T/R switching (ZTE) and gradient ramping (UTE).

Here we modify and apply for the first time a robust and computationally efficient missing point and phase estimation algorithm originating in the solid state NMR community for ZTE imaging sequences.

Patient motion continues to be a major problem in MRI. We propose and validate a novel deep learning approach for the correction of large movements in brain MRI. Training pairs were generated using in-house MRI data of high quality, and simulated images with artifacts based on real head movements. The images predicted by the proposed DL method from motion-corrupted data have improved image quality compared with the original corrupted images in terms of a quantitative metric and visual assessment by experienced readers.

We have previously shown the potential efficacy of a magnetic resonance fingerprinting (MRF) residual approach for human cerebral cortex parcellation. However, the classical Bloch equations, commonly used for MR signal simulation in the MRF dictionary, may not accurately describe the complex effect of the ensemble of microarchitectonic components of the gray matter tissue on MRF signal. This work benefitted from the more flexibility of the extended time-fractional order Bloch equations to improve MRF dictionary fitting accuracy. We demonstrated that the time-fractional order parameter α potentially associates with the effect of interareal architectonic variability, hypothetically leading to more accurate cortical parcellation.

Due to the capability of fast multi-parametric quantitative imaging, magnetic resonance fingerprinting has become a promising quantitative magnetic resonance imaging (QMRI) approach. However, the highly undersampled and noise-contaminated k-space data will cause critical spatial artifacts, which subsequently lead to inaccurate estimation of the quantitative parameters. In this paper, we introduce a novel framework based on structured low-rank approximation and subspace modeling to recover temporal MRF data from its highly undersampled and noisy Fourier coefficients.

In this work, we investigate the influence of eddy currents on hybrid state free precession (HSFP) sequences with 3D radial (koosh-ball) readout trajectories and propose a simulated annealing-based algorithm to minimize their impact.

The proposed approach of reordering the temporal succession of radial spokes successfully minimized the influence of eddy currents on HSFP experiments. Although these results were shown in the context of a HSFP experiment, we believe that this approach could find a broader application in multi-shot bSSFP sequences. 

Super-resolution using deep learning has been successfully applied to camera imaging and recently to static and dynamic MRI. In this work, we apply super-resolution to the generation of high-resolution real-time MRI from low resolution counterparts in the context of human speech production. Reconstructions were performed using full (ground truth) and truncated zero-padded k-space (low resolution). The network, trained with a common 2D residual architecture, outperformed traditional interpolation based on PSNR, MSE, and SSIM metrics.  Qualitatively, the network reconstructed most vocal tract segments including the velum and lips correctly but caused modest blurring of lip boundaries and the epiglottis.

The main focus of this work is to introduce an unsupervised deep generative manifold model for the alignment and joint recovery of the slices in free-breathing and ungated cardiac cine MRI. The main highlights are

(1)   the ability to align multi-slice data and capitalize on the redundancy between the slices.

(2)   The ability to estimate the gating information directly from the k-t space data.

(3)   The unsupervised learning strategy that eliminates the need for extensive training data.

The joint recovery facilitates the acquisition of data from the whole heart in around 2 minutes of acquisition time.

Deep learning image reconstruction algorithms often suffer from model mismatches when the acquisition scheme differs significantly from the forward model used during training. We introduce a Generalized Stein's Unbiased Risk Estimate (GSURE) loss metric to adapt or fine-tune the network to the measured k-space data, thus minimizing the impact of model misfit. Unlike current methods that rely on the mean square error in k-space, the proposed metric accounts for noise in the measurements. This makes the approach less vulnerable to overfitting, thus offering improved reconstruction quality compared to schemes that rely on mean-square error.

Compressed Sensing theory is often applied to accelerate the acquisition of multi-contrast MR images. When highly undersampled, CS-MRI suffers from non-negligible reconstruction error. Here we propose an unrolled iterative deep-learning model to further utilize the group sparsity property for multi-contrast MRI reconstruction at high acceleration factor, named Joint-ISTA-Net, to reduce reconstruction error and aliasing. Our method adds a joint-shrinkage-thresholding model into ISTA-Net to generate a better reconstruction for multi-contrast image pairs. Experiments show the effectiveness of the proposed strategy.