1351039 - Repeatability of Cardiac MR Fingerprinting over Multiple Days in the ISMRM/NIST MRI System Phantom and Healthy Subjects

In Magnetic Resonance Fingerprinting (MRF) (1), Bloch equation simulations are used to mitigate the effects of confounding factors on T₁ and T₂ measurements (2), which may lead to improved accuracy and repeatability. While studies have assessed the repeatability of brain MRF techniques (3,4), similar validation has not been performed for cardiac MRF. This work characterizes the repeatability of cardiac MRF over multiple days in a standardized phantom and healthy subjects.

Methods:

Phantom Study: Experiments were performed using the ISMRM/NIST MRI system phantom (5) at 1.5T (Sola, MAGNETOM Siemens). To evaluate repeatability across heart rates, scans were performed in one session using heart rates ranging from 40 to 120bpm (step size 20bpm). To evaluate intersession repeatability, data were collected during 9 sessions over 8 months (all with 60bpm). MRF maps were collected using a 15-heartbeat sequence with ECG triggering (192x192 matrix, 300x300mm² FOV, 8mm slice) and a deep image prior reconstruction (4). MOLLI and T₂-prepared bSSFP maps with matched resolution were collected for comparison. Reference values were obtained using inversion recovery and spin echo sequences. Repeatability was assessed using the coefficient of variation (CV), defined as the standard deviation normalized by the reference. In Vivo Study: Four healthy subjects underwent repeated scans (3 subjects during 2 sessions, and 1 subject during 3 sessions) with at least one week between scans using MRF, MOLLI, and T₂-prep bSSFP at a mid-ventricular slice. The repeatability of T₁ and T₂ estimates in the left ventricular septum was assessed using CV values.


Results: As shown in Figures 1A-1B, MRF had improved repeatability in the phantom compared to conventional mapping across different heart rates. The average CV for T₁ was 0.9% (MRF) vs 3.0% (MOLLI), and T₂ 2.6% (MRF) vs 7.9% (T₂-prep bSSFP). Figures 1C-1D show repeatability over multiple days in the phantom. The average CV for T₁ was 2.2%(MRF) vs 4.4% (MOLLI); excluding three vials with T₁<200ms yielded 1.7% (MRF) vs 1.6% (MOLLI). The average CV for T₂ was lower for MRF (3.6%) than T₂-prep bSSFP (7.3%). As shown in Figures 2A-2B, MOLLI T₁ and both MRF T₁ and T₂ agreed with the reference, while T₂ was overestimated using T₂-prep bSSFP. Figures 2C-2F show T₁ and T₂ measurements from individual sessions. In vivo repeatability results are shown in Figure 3. MRF had similar repeatability to conventional sequences with an average CV over all subjects of 0.7% (MRF) vs 0.9% (MOLLI) for T₁ and 2.4% (MRF) vs 2.0% (T₂-prep bSSFP) for T₂.


Conclusion:

In phantom experiments, MRF was less susceptible to heart rate than conventional sequences. Over multiple days, MRF had similar repeatability as MOLLI for T₁ and improved repeatability compared to T₂-prep bSSFP. In vivo repeatability was similar among MRF and conventional techniques in 4 healthy subjects. Future work will include validation in additional subjects and assessment of inter-scanner and multisite reproducibility.