Multiscale white- and gray-matter diffusion abnormalities and their associations with serum ammonia in hepatic encephalopathy: a free-water elimination diffusion MRI study.

BACKGROUND: Hepatic encephalopathy (HE) involves complex brain edema, but conventional diffusion models may be confounded by extracellular water and cannot readily distinguish tissue-specific abnormalities from fluid-related effects. This study used a free-water elimination (FWE) diffusion tensor imaging framework to characterize multiscale white- and gray-matter (GM) diffusion abnormalities in HE and to examine their associations with serum ammonia and clinical measures. METHODS: Patients with HE and healthy controls (HCs) were prospectively enrolled in a cross-sectional observational case-control study. FWE-DTI was used to derive conventional diffusion metrics (FA, MD, AD, and RD), tissue-specific metrics (FAt, MDt, ADt, and RDt), and the fractional volume of free water (FW). White-matter (WM) and cortical GM alterations were evaluated using tract-based spatial statistics (TBSS) and GM-based spatial statistics (GBSS), respectively. Conventional peak width of skeletonized mean diffusivity (PSMD) quantified global WM diffusivity heterogeneity. Within the HE group, voxel-wise multiple linear regression models assessed associations of diffusion metrics with serum ammonia, MELD, PHES, and MMSE, adjusting for age, sex, and TIV. TFCE with 5,000 permutations and FWER correction within each statistical map was applied. RESULTS: Compared with HCs, patients with HE showed elevated PSMD and widespread diffusion abnormalities across the WM and cortical GM skeletons. Increased FW was observed in the WM skeleton. Within the HE group, higher serum ammonia was associated with lower AD, ADt, and FW in WM and with lower AD, MD, RD, ADt, MDt, and RDt in cortical GM. Exploratory analyses identified positive associations between PHES and selected white- and GM diffusion metrics and between MMSE and selected WM diffusion metrics; no significant MELD-related associations were detected. All reported voxel-wise associations survived TFCE-based FWER correction. CONCLUSIONS: The findings indicate coexisting white- and GM diffusion abnormalities in HE and distinct within-HE associations between serum ammonia and selected diffusion metrics. FWE-derived metrics provide complementary estimates of free-water fraction and tissue-specific diffusion, but their cellular basis, temporal evolution, and causal interpretation remain uncertain. These findings require validation in larger longitudinal cohorts using multi-shell diffusion MRI and complementary biological assessments.