Temporal-resolution enhanced cardiac cine-MRF for simultaneous T1, T2 mapping and left-ventricular function assessment at 0.55 T

Background: Cardiac Magnetic Resonance Fingerprinting (cMRF) enables co-registered T1 and T2 mapping within a single scan. Free-running cMRF (FR-cMRF) has been proposed for simultaneous T1, T2, and cine imaging at 1.5/3T^1,2, but has not yet been demonstrated at 0.55T, a field strength offering reduced cost and improved accessibility³. We recently reported a preliminary FR-cMRF research sequence at 0.55T⁴. In this study, we extended it to higher temporal resolution for combined T1, T2, and cine imaging and evaluate its performance.

Methods: FR-cMRF uses continuous radial tiny-golden angle bSSFP readouts with sinusoidal flip-angle variation (10°–[30, 50, 70]°). An inversion pulse is applied every 5 lobes and 6 T2-preps every 3 lobes. Data are retrospectively assigned to 14 cardiac phases with exponential soft-gating⁵, followed by motion-resolved low-rank inversion (LRI)⁶ reconstruction with HD-PROST⁷ regularization, producing T1/T2/M0 maps. Synthetic cine contrasts are generated via the bSSFP equation⁸, cardiac motion is estimated via non-rigid registration^9,10. The obtained motion fields are interpolated to 28 phases and incorporated in a motion-corrected (MC) reconstruction (LRI with HD-PROST). MC maps and cine images are obtained via dictionary matching.



Validation was performed on a standardized phantom and 8 healthy subjects (29±3yo, 6 male). Scans were performed on a 0.55T MRI scanner (MAGNETOM Free.Max, Siemens) with a 15-channel coil (FOV=256x256mm², voxel-size=2x2x10mm³, TE/TR=3.9/7.8ms, IR-delays=20ms, T2prep-durations=[45,55,65,45,55,65]ms, 2000 radial spokes, scan time 16s). Spin-echo (SE) served as a reference for phantom T1/T2, while T1-MOLLI and T2-bSSFP were used for in-vivo scans. Single-slice left-ventricle (LV) function was assessed in comparison to conventional cine imaging.

Results: In the phantom experiment, FR-cMRF T1/T2 maps (averaged across all cardiac phases) and conventional mapping techniques (MOLLI and T2-bSSFP) showed good agreement (R² >0.99 for T1 and R² >0.97 for T2) with the SE references (Fig.1). Six-AHA segment¹¹ average FR-cMRF T1/T2 diastolic values across all volunteers were 764±31 ms and 64±5 ms, with biases of 23% and 20%, respectively, versus conventional mapping (Fig.2). T1, T2 and cine images for different cardiac phases show good agreement with conventional MOLLI, T2-bSSFP and cine images in terms of image quality and motion dynamics (Fig.3a-b). LV function assessment shows a mean bias of -0.16% and a high degree of correlation between measurements (Pearson r>0.7) (Fig.3c).

Conclusion: The proposed FR-cMRF sequence allows for simultaneous T1, T2 mapping and LV function assessment in a single 16s scan at 0.55T. Good T1-T2 accuracy and precision was observed in phantom, while high precision was observed against in-vivo conventional techniques. Good LV function quantification agreement was observed against conventional cine. Further work includes evaluation on a larger cohort of healthy subjects and patients with suspected cardiovascular disease.

Figure 1. T1-T2 phantom experiments. (A) T1 maps obtained via conventional T1-MOLLI and the proposed FR-cMRF are compared against IR-SE reference maps, showing a good agreement for both approaches against the reference (R²>0.98 and R²>0.99, for T1-MOLLI and FR-cMRF, respectively). (B) T2 maps obtained via conventional T2-bSSFP and the proposed FR-cMRF, showing a high correlation for both approaches against the reference (R²>0.99 and R²>0.97, respectively). A T2 underestimation was observed with FR-cMRF for T2 values outside of the cardiac range at 0.55T (T2>100 ms), consistent with previous reports.
Figure 2. In-vivo regional assessment of the proposed FR-cMRF approach against conventional maps. Average mean/spatial variability values for T1-MOLLI and T2-bSSFP maps are 619±43 ms and 53±5.8 ms, respectively. The mean/spatial variability T1 and T2 values for the proposed FR-cMRF sequence are 764±31 ms and 64±5 ms, respectively, showing higher quantification precision for both relaxation parameters.
Figure 3. Free-running cardiac MRF In-vivo experiment. (A) FR-cMRF T1 and T2 maps are shown across several cardiac phases alongside with T1-MOLLI and T2-bSSFP maps acquired at diastole, for a representative healthy subject. (B) Reference bSSFP cine images are compared for 8 cardiac phases against the synthetic FR-cMRF cine images generated with the proposed method. (C) Bland-Altman plot shows a mean bias of -0.16% for ejection fraction estimation, with a high Pearson correlation index (r > 0.7) and a p-value=0.92, showing no relevant statistical differences between both sets of measurements.