Assessment of Different Strategies for ROI-Based Measurements of Proton Density Fat Fraction in NAFLD

Abstract Aims: MRI-PDFF is a promising non-invasive method of quantifying liver fat. This study assessed the reproducibility and repeatability of MRI-based PDFF in NAFLD patients, the consistency between different ROI sampling strategies, as well as the learning time burden of PDFF measurement. Methods: 90 NAFLD patients and 30 healthy people were enrolled in this cross-sectional study. Three reviewers measured liver PDFF by circling ROIs in four segments, nine Couinaud segments, and a random place on the PDFF map, respectively. Bland-Altman analysis and ICC were used to evaluate the inter- and intra-reviewer agreements for liver PDFF. Results: Circling ROIs within nine Couinaud segments had the narrowest limit of agreement (LOA) for PDFF measurements. For all 120 subjects, the intra-reviewer agreement had mean LOA of ±0.9% for three strategies, while the inter-reviewer agreement had mean LOA of ± 0.93%. We additionally determined that inter-group consistency increased with decreasing liver fat content for all ROI-circling strategies. The CUSUM learning curve shows that after 17 cases, beginners circled the ROIs more efficiently and accurately. Conclusions: Researchers and clinicians can circling one random ROI for mass screening of patients with NAFLD. We suggest training beginners perform at least 17 PDFF maps to acquire reliable data.