Coronary arteries remain challenging structures to image using MRI due to their small caliber, tortuous course, and continual motion. Current techniques use prospective ECG and navigator gating to suppress cardiac and respiratory motion artifacts, respectively. However, these strategies only accept data acquired within a narrow window in the cardiac and respiratory cycles, resulting in low scanning efficiency, prolonged scan time, and time-consuming setup. Several recent works proposed to increase the respiratory gating efficiency through respiratory motion correction, which significantly reduces scan time and largely eliminates the scan time uncertainty [1-3]. Early efforts to apply this concept on cardiac motion have also been proposed [4-5]. In this work, we developed a new cardiac motion correction framework based on 4D coronary MRA [6], with retrospectively selected cardiac acceptance window, non-rigid motion registration, and iterative, motion corrected reconstruction.

MR data was collected using a contrast-enhanced spoiled gradient echo sequence with 3D radial trajectory and retrospective cardiac and respiratory self-gating, from which 16 cardiac phases were reconstructed with affine respiratory motion correction. Next, early and mid-diastolic phases are selected, and registered to a common reference phase using a symmetric diffeomorphic model [7]. Then, motion-corrected reconstruction is accomplished by inverting the encoding operator that includes sensitivity encoding and warping the target image to different cardiac phases [8]. Briefly, the unknown image $$$x$$$ is reconstructed by solving the following optimization problem:

$$\hat{x}=argmin\left\{|Ex-y|_2^2+\lambda|TV(x)|_1\right\}$$

where $$$E$$$ is the encoding operator that maps a motion-free image $$$x$$$ to the multi-channel, multi-cardiac phase k-space data $$$y$$$, $$$TV()$$$ is the spatial total variation (TV) operator, and $$$\lambda$$$ is the regularization weight. The forward operator was implemented as follows:

$$y_{channel,phase}=FT[S_{channel}*T_{phase}^{-1}(x)]$$

where $$$FT$$$ is non-uniform Fourier Transform, $$$S$$$ is the self-calibrated sensitivity map, and $$$T^{-1}$$$ is the spatial warp from the reference to a particular phase. The backward operator was implemented as follows:

$$x=\sum_{channel,phase}S_{channel}^**T_{phase}[FT^{-1}(y_{channel,phase})]$$

where $$$T$$$ is the spatial warp from a particular phase to the reference. The optimization problem was solved using a nonlinear conjugate gradient (CG) solver written in MATLAB.

Healthy subjects (N=11) were scanned using a clinical 3T scanner (Verio, Siemens) with following parameters: TR/TE=6.0/3.7 ms, flip angle = 15°, FOV = (320 mm)³, matrix size = 320³, number of lines = 99,994, scan time = 10 minutes, contrast enhancement with a 0.20 mmol/kg Gd-BOPTA (MultiHance, Bracco) injected at 0.3 mL/s before acquisition. Three images were reconstructed from each dataset: mid-diastole, motion free window (Short), extended window without motion correction (Long), and extended window with motion correction (Long_Moco). The scan efficiency, coronary sharpness, and apparent signal-to-noise ratio (aSNR) werecompared using paired Student’s t-test with a significance level of 0.05.

The mean aSNR for Short, Long, and Long_Moco were 11.23 ­­± 2.90, 12.03 ± 3.09, and 12.72 ± 4.17, respectively. Including the early and mid-diastole phases into the reconstruction significantly improved aSNR compared with the mid-diastole quiescent period only (P=0.045 and P=0.025, with and without motion correction, respectively). The mean coronary sharpness for Short, Long and Long_Moco were 0.21 ± 0.02, 0.20 ± 0.03, and 0.24 ± 0.04, respectively. Long_Moco significantly improved coronary sharpness compared with both Short and Long (P=0.008 and P<0.001, respectively). The mean scan efficiencies for the short and long cardiac window were 14.8% ± 4.0% and 35.2% ± 5.5%, respectively. The numbers are summarized in Fig. 1. Fig. 2 shows images from two example subjects. Only accepting data from a typical mid-diastole window well suppresses motion artifacts but shows considerable residual streaking due to undersampling. Extending the acceptance window to early and mid-diastole increases aSNR but introduces blurring due to cardiac motion. The proposed cardiac motion correction technique effectively suppresses motion blurring while maintaining high aSNR.In this work, we proposed to significantly increase the scanning efficiency of whole-heart coronary MRA by combining early and mid-diastole phases from 4D datasets, and suppress the cardiac motion through non-rigid motion registration and iterative reconstruction. In vivo studies on 11 healthy volunteers showed that the proposed method significantly improved aSNR and coronary sharpness over both the mid-diastole reconstruction and reconstruction from the extended window without motion correction. A higher scan efficiency may be used to improve the image quality given a fixed imaging time, as demonstrated in this study, or to allow the scan time to be shortened, which warrants future investigation.The proposed cardiac motion correction framework significantly increased the scan efficiency of whole-heart coronary MRA, and improves image quality given the same available data. Future works will be focused on further optimizing the motion correction algorithm, and explore the potential to further scan time reduction.