0508

The key to extremely accelerated model-based quantitative first-pass perfusion cardiac MRI
Teresa M Correia1, Torben Schneider2, and Amedeo Chiribiri1
1School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2Philips Healthcare, Guildford, United Kingdom

Synopsis

First-pass perfusion cardiac MR (FP-CMR) is one of the methods of choice for evaluating myocardial ischemia. Moreover, quantitative FP-CMR methods that provide pixel-wise quantitative myocardial perfusion maps are increasingly being applied as an alternative to visual inspection. Recently, a DIRect QuanTitative (DIREQT) FP-CMR model-based reconstruction has been proposed to directly estimate myocardial perfusion maps and highly accelerate FP-CMR scans. Here, DIREQT is combined with the idea of view-sharing and KEYhole imaging (DIREQT-KEY) to improve DIREQT reconstructions, particularly from extremely accelerated FP-CMR acquisitions. DIREQT-KEY directly generates high-quality quantitative myocardial perfusion maps from less than 4 radial spokes per time frame.

Introduction

First-pass perfusion cardiac MR (FP-CMR) is an emerging non-invasive tool for evaluating coronary heart disease (CHD). Recently, FP-CMR methods that provide pixel-wise quantification of myocardial perfusion have gained increased interest as an alternative to visual inspection.1-3 Typically, these methods indirectly generate quantitative myocardial perfusion maps by first reconstructing individual dynamic contrast-enhanced images, which are then converted to contrast agent concentration and, finally, tracer-kinetic (TK) modeling is used to generate quantitative maps. However, FP-CMR time frames must be acquired in real-time to capture the rapid passage of a contrast agent bolus through the heart, and hence, the spatial resolution and coverage of the heart is compromised.4,5 Undersampled reconstruction methods have been proposed to moderately accelerate FP-CMR acquisitions as a means to improve spatial resolution.6-9 Recently, a DIRect QuanTitative (DIREQT) FP-CMR reconstruction method has been proposed to directly estimate quantitative TK maps from highly undersampled acquisitions, by exploiting the redundancy of spatial information between time frames.10 However, some TK models only use the key frames of contrast arrival for quantitative analysis. Nonetheless, high spatial frequencies from the k-space periphery of the non-key frames (or additionally acquired frames) may contain useful information that could be shared to improve DIREQT reconstructions from highly accelerated FP-CMR data. In this work, DIREQT is combined with the idea of view-sharing and KEYhole imaging11 (DIREQT-KEY) to improve the quality of the TK maps, particularly from extremely accelerated FP-CMR acquisitions. The proposed method was tested in four patients with suspected CHD.

Methods

Acquisition Rest FP-CMR fully-sampled acquisitions were performed in four patients with suspected CHD (ethically approved; informed written consent obtained) using a dual-bolus technique with 0.0075+0.075mmol/kg Gadobutrol (Gadovist; Bayer, Germany). Patients 1&2 were scanned on a 3T Philips Achieva scanner and patients 3&4 were scanned on a 1.5T Philips Ingenia scanner (Philips Healthcare, Netherlands). A saturation-recovery turbo field echo (TFE) sequence was used to acquire one short-axis slice in free-breathing using the following parameters: FOV=320x320mm2, resolution=2.8x2.8mm2, slice thickness=10mm, TS/TR/TE= 120/1.96/0.93ms, flip angle=15deg, acquisition window=224-228ms, total acquisition time=1min20s, dynamic frames=90-125, and contrast agent relaxivity=5.0 L/mmol·s. The AIF was found using a region of interest in the left ventricle. The fully-sampled dynamic images were used to estimate the frame-to-frame translational motion. Then, translational motion correction was performed directly in k-space to generate motion-corrected datasets.

Reconstruction Radial (k-t)-space sampling was used to generate 40x (3 spokes), 60x (2 spokes) and 120x (1 spoke) undersampled datasets, which were reconstructed using DIREQT and DIREQT-KEY (Fig.1). The DIREQT method directly estimates TK parameter maps from the measured multicoil (k-t)-space FP-CMR data. This is achieved by solving the following optimization problem: $$$\widehat{\textbf{M}}=\underset{\textbf{M}}{\arg\min}\parallel\textbf{d}-f(\textbf{M})\parallel_2^2$$$, where $$$\textbf{M}$$$ are the TK parameters maps (e.g. KTrans and vp of the Patlak model), $$$\textbf{d}$$$ is the undersampled (k-t)-space data and $$$f$$$ is the forward model (indicated by the small red arrows in Fig. 1a). A limited memory BFGS quasi-Newton method was used to solve this nonlinear inverse problem. Three different versions of DIREQT-KEY were tested (Fig. 1b): 1) a central “keyhole” keeps the acquired (k-t)-space data and the peripheral data from the non-key frames are shared; 2) same as 1, but the central region also shares the weighted down (k-t)-space data from the non-key frames; 3) same as 1, but exponentially weighted down data from adjacent times frames is used in each “keyhole". In the latter, most weights will be closer to zero, except for the 4 closest time frames.

Results and Discussion

Figure 2 shows the DIREQT and DIREQT-KEY reconstructions obtained for patient 1. The structural similarity index (SSIM) indicates that all three versions of DIREQT-KEY generate TK maps with superior image quality compared to the DIREQT method. The highest SSIM was achieved with DIREQT-KEY version 2, which indicates a better agreement with the reference. Hence, the remaining analysis was done using DIREQT-KEY version 2. Figure 3 shows the quantitative maps obtained for patients 2-4 using the DIREQT and DIREQT-KEY methods. Patient 2 had myocardial infarction and patient 4 hypertrophic cardiomyopathy. Patients 1 and 3 had no abnormalities on CMR. The perfusion defect in Patient 2 was identified in both DIREQT and DIREQT-KEY maps and corresponds to an area of myocardial scarring. DIREQT-KEY improvements are more noticeable for Patient 1&4 since these had more extra frames. However, additional time frames could be acquired to improve DIREQT-KEY maps (Fig. 4-5). DIREQT and DIREQT-KEY provide high-quality maps even at extremely high acceleration rates (Fig. 3-5). However, DIREQT-KEY provides the highest (SSIM) quality TK maps (Fig. 5). Spatial regularization could be used to reduce noise amplification at high accelerations.

Conclusion

The proposed DIREQT-KEY, combines DIREQT, a model-based reconstruction approach, with view-sharing and keyhole imaging. Similar to DIREQT, DIREQT-KEY allows extremely high FP-CMR scan acceleration factors (< 4 radial spokes) and quantitative imaging. However, DIREQT-KEY provides higher quality TK maps. Moreover, DIREQT-KEY works with existing (k,t)-sampling, dual-bolus and dual-sequence acquisitions strategies. These aspects enhance the clinical applicability of the method. In future studies, the DIREQT-KEY method will be evaluated in a large cohort of patients with suspected CHD using prospectively undersampled acquisitions. These studies will also aim to achieve much higher spatial resolution and coverage, and hence, greater diagnostic accuracy.

Acknowledgements

This work was supported by the Wellcome/EPSRC Centre for Medical Engineering [WT 203148/Z/16/Z].

References

1. Kellman P, Hansen MS, Nielles-Vallespin S, et al. Myocardial perfusion cardiovascular magnetic resonance: optimized dual sequence and reconstruction for quantification. J Cardiovasc Magn Reson. 2017;19(1):43.

2. Hsu LY, Jacobs M, Benovoy M, et al. Diagnostic performance of fully automated pixel-wise quantitative myocardial perfusion imaging by cardiovascular magnetic resonance. JACC Cardiovasc Imaging. 2018;11(5):697-707.

3. Villa ADM, Corsinovi L, Ntalas I, et al. Importance of operator training and rest perfusion on the diagnostic accuracy of stress perfusion cardiovascular magnetic resonance. J Cardiovasc Magn Reson. 2018;20(1):74.

4. Kellman P, Arai AE. Imaging sequences for first pass perfusion - a review. J Cardiovasc Magn Reson. 2007;9(3):525-37.

5. Fair MJ, Gatehouse PD, DiBella EV, Firmin DN. A review of 3D first-pass, whole-heart, myocardial perfusion cardiovascular magnetic resonance. J Cardiovasc Magn Reson. 2007;9(3):525-37.

6. Otazo R, Kim D, Axel L, Sodickson DK. Combination of compressed sensing and parallel imaging for highly accelerated first-pass cardiac perfusion MRI. Magn Reson Med. 2010;64(3):767-76.

7. Vitanis V, Manka R, Giese D, et al. High resolution three-dimensional cardiac perfusion imaging using compartment-based k-t principal component analysis. Magn Reson Med. 2011;65(2):575-87.

8. Lingala SG, Hu Y, DiBella E, Jacob M. Accelerated dynamic MRI exploiting sparsity and low-rank structure: k-t SLR. IEEE Trans Med Imaging. 2011;30(5):1042-54.

9. Chen X, Salerno M, Yang Y, Epstein FH. Motion-compensated compressed sensing for dynamic contrast-enhanced MRI using regional spatiotemporal sparsity and region tracking: block low-rank sparsity with motion-guidance (BLOSM). Magn Reson Med. 2014;72(4):1028-38.

10. Correia T, Schneider T, Chiribiri A. Model-Based Reconstruction for Highly Accelerated First-Pass Perfusion Cardiac MRI. In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2019. Lecture Notes in Computer Science, Springer. 2019;11765.

11. van Vaals J, Brummer M, Dixon W, et al. ‘Keyhole’ method for accelerating imaging of contrast agent uptake. J Magn Reson Imaging 1993;3: 671-5.

Figures

Figure 1. a) The indirect method consists of three sequential steps (gray arrows), whereas DIREQT estimates TK maps (e.g. KTrans and vp) directly from (k-t)-space data (long red arrow). DIREQT-KEY combines DIREQT with keyhole and view-sharing imaging: c) version 1 - central “keyhole” keeps the acquired data and the peripheral data from b) non-key frames (not used for quantification) is shared; d) version 2 - central region also shares the weighted down data from non-key frames; e) version 3 - exponentially weighted down data from adjacent times frames is used in each “keyhole”.

Figure 2. Quantitative (KTrans) maps obtained from 40x (3 spokes), 60x (2 spokes) and 120x (1 spoke) undersampled patient data using DIREQT and three versions of DIREQT-KEY. The reference image (fully-sampled) is displayed for comparison. The structural similarity index (SSIM), which is displayed in each subfigure, indicates a good agreement between the reference and DIREQT(-KEY) maps. However, DIREQT-KEY provides higher quality maps (higher SSIM) compared to DIREQT. The highest SSIM was achieved with DIREQT-KEY version 2, which indicates a better agreement with the reference.

Figure 3. Quantitative (KTrans) maps obtained from fully-sampled (reference), 40x (3 spokes), 60x (2 spokes) and 120x (1 spoke) undersampled patient data using DIREQT and DIREQT-KEY (version 2). Patients 3 had no abnormalities on CMR, patient 2 had myocardial infarction (yellow arrow) and patient 4 hypertrophic cardiomyopathy. DIREQT provides quantitative maps that show good visual agreement with the reference maps for all patients. DIREQT-KEY can be used to further improve reconstructions and achieve higher quality quantitative maps from just one radial spoke per time frame.

Figure 4. Quantitative (KTrans) maps obtained from 60x (2 spokes) and 120x (1 spoke) undersampled patient data using DIREQT-KEY (version 2). DIREQT-KEY combines DIREQT with keyhole and view-sharing imaging, where a central “keyhole” keeps the acquired (k-t)-space data and the missing peripheral data can be obtained, for example, from non-key frames (those not used for quantification). Prolonging FP-CMR acquisitions can be useful to obtain even better quality quantitative maps. In this example, a non-key frame was repeated to make a total of 250 time frames (125-160 extra frames).

Figure 5. Structural similarity index (SSIM) a) between the reference and TK maps obtained with DIREQT and DIREQT-KEY (version 2) for all patients. Both DIREQT and DIREQT-KEY allow for extremely high acceleration factors. However, the highest SSIM values were achieved with DIREQT-KEY, which indicates a better agreement with the reference. b) SSIM between the references TK maps and DIREQT-KEY with extra time frames. Higher quality maps (higher SSIM) can be achieved by prolonging the FP-CMR scan and using the additional peripheral data in the DIREQT-KEY reconstruction.

Proc. Intl. Soc. Mag. Reson. Med. 29 (2021)
0508