Stratifying non-alcoholic fatty liver disease using diffusion-weighted imaging-based habitat imaging: A rabbit model study.

Abstract

PURPOSE: Habitat imaging integrated multiparametric biomarkers, enabling precise quantification and visualization of hepatic tissue characteristics. This work evaluated DWI-based habitat imaging for stratifying Non-alcoholic Fatty Liver Disease (NAFLD) stages to enable early detection of Nonalcoholic Steatohepatitis (NASH). METHODS: Using existing DWI data from 32 male New Zealand rabbits randomized to high-fat/cholesterol or standard diets (NAFLD model), we applied a novel habitat imaging framework integrating mono-exponential diffusion-weighted imaging (DWI), intravoxel incoherent motion (IVIM), stretched exponential model (SEM), and diffusion kurtosis imaging (DKI) models. This identified four hepatic habitats reflecting perfusion, diffusion heterogeneity, and microstructural complexity. Unlike prior diffusion analyses, habitat fractions (H- H) and heterogeneity (HH) indices were derived via spatial multiparametric pattern recognition. Diagnostic performance for NASH was assessed by AUC. RESULTS: NASH was characterized by a significantly lower fraction of habitat 1 (H) and higher fraction of habitat 4 (H) (p&#xa0;<&#xa0;0.05). Habitat heterogeneity (HH) was markedly reduced in NASH versus normal histology and simple steatosis (p&#xa0;<&#xa0;0.05). For distinguishing NASH, Hdemonstrated a numerically higher AUC (0.931) compared to H&#xa0;+&#xa0;H(AUC&#xa0;=&#xa0;0.922) and perfusion fraction (f, AUC&#xa0;=&#xa0;0.909). However, the DeLong test confirmed no statistically significant differences among these AUC values. CONCLUSION: DWI habitat imaging is a promising tool for NAFLD monitoring, detecting tissue heterogeneity with diagnostic performance comparable to conventional diffusion metrics.