Feasibility of 3D MRI fingerprinting for rapid knee cartilage T1, T2, and T1ρ mapping at 0.55T: Comparison with 3T

Low‐field strength scanners present an opportunity for more inclusive imaging exams and bring several challenges including lower signal‐to‐noise ratio (SNR) and longer scan times. Magnetic resonance fingerprinting (MRF) is a rapid quantitative multiparametric method that can enable multiple quantitative maps simultaneously. To demonstrate the feasibility of an MRF sequence for knee cartilage evaluation in a 0.55T system we performed repeatability and accuracy experiments with agar‐gel phantoms. Additionally, five healthy volunteers (age 32 ± 4 years old, 2 females) were scanned at 3T and 0.55T. The MRI acquisition protocols include a stack‐of‐stars T1ρ‐enabled MRF sequence, a VIBE sequence with variable flip angles (VFA) for T1 mapping, and fat‐suppressed turbo flash (TFL) sequences for T2 and T1ρ mappings. Double‐Echo steady‐state (DESS) sequence was also used for cartilage segmentation. Acquisitions were performed at two different field strengths, 0.55T and 3T, with the same sequences but protocols were slightly different to accommodate differences in signal‐to‐noise ratio and relaxation times. Cartilage segmentation was done using five compartments. T1, T2, and T1ρ values were measured in the knee cartilage using both MRF and conventional relaxometry sequences. The MRF sequence demonstrated excellent repeatability in a test–retest experiment with model agar‐gel phantoms, as demonstrated with correlation and Bland–Altman plots. Underestimation of T1 values was observed on both field strengths, with the average global difference between reference values and the MRF being 151 ms at 0.55T and 337 ms at 3T. At 0.55T, MRF measurements presented significant biases but strong correlations with the reference measurements. Although a larger error was present in T1 measurements, MRF measurements trended similarly to the conventional measurements for human subjects and model agar‐gel phantoms. Low‐field scanners are cheaper alternatives that can improve accessibility to MRI while bringing challenges such as lower signal‐to‐noise ratio efficiency (SNRe). Using MRI fingerprinting (MRF), phantom and volunteers were scanned in a 0.55T scanner to assess the SNRe and accuracy. Measurements showed excellent repeatability and correlation with reference measurements.