Hui Xue1, Ding Yu2, Saurabh Shah3,
Christoph Guetter1, Marie-Pierre Jolly1,
1Imaging &
Visualization, Siemens Corporate Research, Princeton, NJ, United States; 2The
Ohio State University, Columbus, OH, United States; 3CMR Research
& Development, Siemens Medical Solutions USA, Inc., Chicago, IL, United
States; 4National Heart, Lung & Blood Institute, National
Institutes of Health, Bethesda, MD, United States
Recent development in cardiac MR shows that improved SNR can be achieved by averaging the motion-corrected free-breathing images using non-rigid image registration. To motion correction induced artifacts, previous studies heuristically exclude some frames from final averaging. However, non-rigid image registration can lead to variant accuracy for different frames and different regions within a frame. We therefore propose a novel image combination algorithm, computing optimal weights for image combination by minimizing the total amount of deformation brought into the averaging. Compared to uniform averaging, our approach achieves good noise suppression and provides better tolerance to artifacts possibly introduced by motion correction.