Spatial characterization of intrahepatic fat deposition across steatotic liver disease subcategories: a multicenter...

Intrahepatic fat accumulation presents spatial heterogeneity, potentially varying across steatotic liver disease (SLD) subcategories. Current whole-liver spatial profiling remains limited. In this study, we systematically evaluated spatial fat distribution patterns in metabolic dysfunction-associated steatotic liver disease (MASLD), alcohol-related liver disease (ALD), and SLD with chronic hepatitis B (SLD&CHB) using a fully automated nnU-Net model for whole-liver segmentation and two-point Dixon MRI-based fat fraction mapping.

This multicenter retrospective study included the analysis of 682 patients with hepatic steatosis (MASLD: 451; ALD: 89; SLD&CHB: 142) who underwent two-point Dixon MRI. The fully automated nnU-Net-based MRI framework was developed with in-phase (IP), out-phase (OP), water, and fat images, and the model performance was quantified using the Dice similarity coefficient (DSC) score. The fat fraction (FF), standard deviation of FF (SD-FF), lobar asymmetry, and periportal-to-peripheral gradients were compared across SLD subcategories and steatosis severity levels (L1: mild, L2: moderate, L3: severe).

The segmentation model achieved: whole-liver DSC: 0.94; left lobe DSC: 0.94; right lobe DSC: 0.96. Compared with ALD and SLD&CHB, MASLD exhibited the highest fat burden (FF: 16.29%) and greatest spatial heterogeneity (SD-FF: 8.22). FF was significantly higher in the right lobe than in the left lobe in MASLD and ALD (both P < .05), whereas no significant lobar difference was observed in SLD&CHB. SD-FF was consistently higher in the left lobe across all SLD subcategories (P < .001), indicating greater heterogeneity of fat distribution in the left lobe. Periportal-to-peripheral gradient analysis revealed peripheral fat enrichment in MASLD and ALD, while the opposite trend in SLD&CHB.

This study developed a fully automated quantification model for the spatial distribution of whole-liver fat. Characterization of the spatial heterogeneity of hepatic fat accumulation provides additional, phenotypically relevant information across SLD subcategories. These findings highlight the clinical utility of spatial heterogeneity as a novel, noninvasive imaging biomarker for subtype identification and personalized disease assessment of SLD. Moreover, this automated framework offers a scalable and objective tool for future longitudinal monitoring.