Deep Image Prior Denoising of Cardiac FISP-MRF T1, T2 and M0 Maps at 0.55T

Motivation: 0.55T scanners offer several potential advantages for quantitative myocardial tissue characterization. However, a challenge at low field is the reduced SNR, which can lead to decreased precision that may compromise the utility of mapping at 0.55T. Goal(s): To combine cardiac MR Fingerprinting (cMRF) with a Deep Image Prior (DIP) to improve mapping precision at 0.55T. Approach: cMRF data were acquired in phantoms and 10 healthy subjects at 0.55T, with maps compared using low-rank and DIP reconstruction techniques. Results: cMRF at 0.55T yielded high-quality maps in all subjects (mean T1 678 +/-24ms, T2 50.4 +/-5.8ms), with DIP yielding improved precision than the low-rank reconstruction. Impact: This study demonstrates the feasibility of FISP-based cardiac MR Fingerprinting for T1, T2, and M0 mapping at 0.55T using a Deep Image Prior reconstruction to suppress noise at low field, outperforming a low-rank MRF reconstruction method.