A Comparison of Tissue Property Values Estimated Using Conventional Cardiac MRF and MT-Cardiac MRF

To evaluate the impact of explicitly modeling Magnetization Transfer (MT) during reconstruction of cardiac Magnetic Resonance Fingerprinting (cMRF) maps on T₁ and T₂ quantification.

cMRF data were reconstructed using two dictionaries: a conventional cMRF dictionary that ignores MT processes, and an MT-cMRF dictionary modeling T₁ and T₂ of the free water pool (T_1f and T_2f, respectively), as well as the bound pool fraction (BPF) as free parameters. Measurements which include MT processes were assessed in simulation and phantom to ensure that T_1f, T_2f, and BPF could accurately be measured using cMRF. cMRF data were then collected in 14 healthy subjects to measure T₁/T_1f, T₂/T_2f, and BPF in normal myocardium. Finally, cMRF data were acquired in five patients with known or suspected myocardial scar at 1.5T, and tissue property maps both ignoring and including MT were generated. Linear discriminant analyses (LDA) were performed to assess differences in tissue property values between scar and remote myocardium.

MT-cMRF showed excellent agreement with ground truth simulations (RMS error = 2.4% ± 3.9%) and phantom measurements (R² = 0.99). Values measured with MT-cMRF in healthy subjects (T_1f = 1137 ± 72 ms, T_2f = 42.8 ± 4.2 ms, BPF = 8.2% ± 1.2%) were significantly different than conventional cMRF (T₁ = 947 ± 75 ms, T₂ = 41.5 ± 3.9 ms). T_1f/T_2f and T_2f/BPF from MT-cMRF demonstrated higher LDA accuracy (82.4% and 86.1%) in discriminating scar from healthy tissue than T₁ and T₂ from conventional cMRF (61.8%).

Incorporating MT modeling into cMRF results in improved T_1f and T_2f quantification compared to conventional modeling and enables BPF measurement. Native MT-cMRF T_1f, T_2f, and BPF values showed differences between scar and healthy myocardium.