Restricted Spectrum Imaging for Diagnosis of High‐Risk Metabolic Dysfunction‐Associated Steatohepatitis

BACKGROUND: Conventional diffusion MRI inadequately characterizes heterogeneous microstructural changes in metabolic dysfunction-associated steatotic liver disease (MASLD). The optimal restricted spectrum imaging (RSI) model and value for high-risk metabolic dysfunction-associated steatohepatitis (MASH) remain unclear. PURPOSE: To optimize RSI modeling and evaluate its performance for high-risk MASH. STUDY TYPE: Prospective cross-sectional study. POPULATION: 96 biopsy-proven MASLD participants (median age, 46 years [IQR: 36-58]; 51 women [53.1%]), including 64 with high-risk MASH. FIELD STRENGTH/SEQUENCE: 3T; fat-suppressed single-shot echo-planar imaging with multi-b-value diffusion-weighted imaging, EPI-based MR elastography, and 3D chemical shift-encoded proton density fat-fraction imaging. ASSESSMENT: Two- to five-compartment RSI models were compared using Bayesian information criterion (BIC). RSI fractions, apparent diffusion coefficient, FIB-4, and MRI-based MASH score (MAST) were evaluated against histopathologic features and high-risk MASH. STATISTICAL TESTS: Spearman correlation, Mann-Whitney U test, logistic regression, receiver operating characteristic analysis, DeLong test, bootstrap internal validation, decision curve analysis (thresholds, 0.20-0.60), and intraclass correlation coefficient (ICC). p < 0.05 was significant. RESULTS: The five-compartment model showed the lowest BIC (1472.30; ΔBIC = 0). RSI measurements showed excellent inter-observer agreement (ICC, 0.955-0.994). RSI-C1 (r = 0.357) and RSI-C2 (r = 0.283) significantly correlated with steatosis; RSI-C2 (r = 0.273) and RSI-C5 (r = 0.282) with inflammation; and RSI-C2 (r = 0.326), RSI-C4 (r = 0.400), and RSI-C5 (r = 0.386) with fibrosis (all p < 0.05). The RSI-C2 + C4 + C5 model yielded an AUC of 0.800 for high-risk MASH, without significant difference versus FIB-4 (AUC = 0.707; p = 0.219) or MAST (AUC = 0.899; p = 0.146). Across thresholds of 0.20-0.60, RSI mean net benefit (0.504) exceeded FIB-4 (0.438) but not MAST (0.559). DATA CONCLUSION: RSI provides compartment-specific microstructural biomarkers associated with MASLD histology and may complement noninvasive tools for high-risk MASH assessment. EVIDENCE LEVEL: 2. TECHNICAL EFFICACY: Stage 2.