3D High SNR Cardiac MRI via Motion-Corrected Averaging of Multi-Heartbeat Acquisitions

Motivation: 3D high-SNR cardiac MRI may be achieved by acquiring undersampled images with low SNR during multiple heartbeats and averaging these volumes after motion correction. Goal(s): To generate 3D high SNR cardiac MRI from free-breathing multi-heartbeat undersampled acquisitions. Approach: We proposed an algorithm that implemented deep-learning based undersampled image reconstruction and deformable motion correction to generate high SNR 3D MRI. Results: For 11 subjects, our approach demonstrated effectiveness in respiratory and cardiac motion correction and generated 3D MR images with SNR 1.7x higher than single heartbeat acquisition and 1.4x higher than that without motion correction. Impact: The proposed approach provides a way for respiratory and cardiac motion correction and enables 3D MRI with high SNR that is required for a wide range of clinical applications.