Improved Liver T1, T2, T2*, and PDFF Mapping at 0.55T Using Rosette MRF with Optimized Sequence Design and Deep Image Reconstruction

Motivation: Quantitative liver imaging at 0.55T suffers from poor precision due to lower SNR and suboptimal acquisitions. Goal(s): Increase the precision of liver MRF T1/T2/T2*/PDFF mapping at 0.55T through sequence optimization and Deep Image Prior (DIP) reconstruction. Approach: The Cramer-Rao Bound was used to design a magnetization-prepared MRF sequence for 2D liver imaging at 0.55T. Raw data were reconstructed using an extension of the DIP framework for multiecho MRF data. Results: Sequence optimization and DIP reconstruction each increased the precision of relaxation time estimation. Combining both approaches reduced the standard deviation by 63%/78%/65% for T1/T2/T2* compared to an unoptimized sequence with a low-rank reconstruction. Impact: Sequence optimization and DIP reconstruction improved the precision of quantitative liver imaging with MRF at 0.55T, potentially enabling liver health assessment in patients who cannot be scanned on higher-field MRI due to accessibility, implantable device, or body size restrictions.