High-quality Four-dimensional Magnetic Resonance Fingerprinting Reconstruction Using Temporal Low-rank-constrained Motion Compensation

Motivation: While motion compensation has been employed in four-dimensional magnetic resonance fingerprinting (4DMRF) reconstruction, suboptimal registration accuracy currently degrades the image quality of motion-resolved tissue maps, precluding its clinical application. Goal(s): Our goal was to develop a high-quality 4DMRF (HQ-4DMRF) reconstruction method. Approach: We proposed a temporal low-rank-constrained groupwise 4D registration algorithm (TLc -G4DReg) to achieve robust and precise inter-phase registration for HQ-4DMRF reconstruction. Results: Results demonstrated that TLc -G4DReg achieved the best performance in comparison with other registration methods used for 4DMRF reconstruction in previous studies. Impact: This study offers a new registration approach that leverages the temporal low-rank property of respiratory motion as prior knowledge. This method has the potential to register extremely under-sampled images groupwisely, enhancing motion-resolved reconstruction, including but not limited to 4DMRF reconstruction.