Investigation of cerebral cortical morphological similarity and network topological abnormalities in hepatic encephalopathy utilizing a morphometric inverse divergence network framework

Background Hepatic encephalopathy (HE) requires objective biomarkers for early diagnosis and mechanistic clarification. This study first integrates the Morphometric Inverse Divergence (MIND) network with graph theory to explore cerebral cortical morphological similarity and topological abnormalities in HE, cirrhotic non-HE (NHE), and healthy control (HC) groups. Methods A total of 31 HE, 30 NHE patients and 30 HCs were enrolled for 3.0T magnetic resonance imaging (MRI) 3D-T1WI scanning. FreeSurfer was used for image preprocessing, and 5 cortical morphological features were extracted based on the Schaefer-400 atlas. MIND networks were constructed via symmetric Kullback–Leibler divergence, graph theory was applied to extract topological properties, and intergroup differences were analyzed by general linear model (GLM). Results Compared with HCs, HE patients exhibited significantly elevated mean MIND values across multiple functional subnetworks, including the visual (VIS; t = 3.629, p = 0.004), default mode (DMN; t = 3.115, p = 0.009), limbic (LMB; t = 2.969, p = 0.009), frontoparietal (FPN; t = 2.917, p = 0.009), and ventral attention (VAN; t = 2.212, p = 0.043) networks. Graph theoretical analysis revealed increased global efficiency (Eglob, t = 2.681, p = 0.0100) and local efficiency (Eloc, t = 2.683, p = 0.010). NHE patients showed mild DMN connectivity enhancement in edge analysis, but no significant differences in subnetwork mean MIND and nodal metrics (all p > 0.05), exhibiting a non-significant transitional trend in efficiency indices between HE and HC groups. Conclusion HE causes abnormal whole-brain cortical morphological covariance networks with enhanced connectivity and efficiency, and NHE has early-stage network alterations. MIND network indices are potential imaging biomarkers for HE diagnosis and monitoring, supplementing the neuropathological mechanism of HE and making up for the limitations of traditional structural covariance network research in HE.