Abstract
Abstract
Purpose
Investigating the feasibility of variable flip angle (VFA) T
1
mapping for quantitative liver imaging at 0.55 T and evaluating the impact of RF spoiling correction and deep learning-based (DL) reconstruction on accuracy.
Methods
A VFA protocol was developed and combined with DL reconstruction to improve quantitative robustness at 0.55 T. Phantom experiments validated T
1
estimates against multi-TI inversion-recovery spin-echo reference measurements. Bloch simulations characterized RF spoiling-induced bias and derived 0.55 T-specific calibration coefficients. Twelve healthy subjects were scanned to assess field-strength-dependent effects of B
1
+
inhomogeneity.
Results
In vivo, DL reconstruction reduced noise-related variability of voxel-wise hepatic T
1
estimates compared to conventional reconstructions in the same subject. At 0.55 T, B
1
+
maps showed the highest transmit homogeneity, with liver flip angle variation of 98.2 ± 2.9%, compared with 94.6 ± 7.4% at 1.5 T and 100.7 ± 9.8% at 3 T. Accordingly, B
1
+
correction had minimal impact on hepatic T
1
distributions at 0.55 T, whereas more pronounced effects were observed at higher field strengths. Phantom experiments demonstrated improved agreement of VFA-derived T
1
values with the IR-SE reference after 0.55 T-specific RF spoiling correction, supporting the accuracy of the corrected VFA approach. Bland–Altman analysis confirmed that RF spoiling correction had a greater impact on phantom T
1
accuracy than B
1
+
correction.
Conclusion
At 0.55 T, RF spoiling and reconstruction strategies are the main determinants of quantitative accuracy, while transmit field inhomogeneity plays a minor role. Combined RF spoiling correction and DL reconstruction enables robust VFA T
1
mapping without B
1
+
correction. Reduced low-field T
1
values improve VFA model conditioning and reduce sensitivity to flip angle deviations, supporting VFA-based T
1
mapping for low-field MRI.