Hepatic encephalopathy (HE) is a common neurological complication in patients with cirrhosis, manifesting as a spectrum of neuropsychiatric symptoms ranging from subtle cognitive impairment to coma (1).

With the aging of the global population, the differential diagnosis between HE and neurodegenerative disorders has become increasingly challenging due to overlapping clinical presentations (2). Traditional diagnostic approaches primarily rely on clinical assessment and blood biochemical markers such as ammonia levels, but these methods are limited by subjectivity and insufficient sensitivity (3).

Among various neuroimaging biomarkers, the globus pallidus (GP) has garnered significant attention in HE research. Magnetic resonance imaging (MRI) studies have demonstrated that 70–90% of HE patients exhibit bilateral symmetric hyperintensity in the GP on T1-weighted images (4). The relationship between this pallidal T1 hyperintensity and plasma ammonia levels remains a subject of debate. While some studies have identified significant associations between pallidal signal intensity and ammonia concentration (5, 6), others have failed to find such a correlation (7). This suggests that pallidal T1 hyperintensity is primarily linked to manganese-related neurotoxicity but may also involve additional mechanisms that are partly distinct from hyperammonemia (8).

While conventional structural MRI captures macroscopic changes, it offers limited sensitivity to microstructural alterations. Neurite Orientation Dispersion and Density Imaging (NODDI) is an advanced diffusion MRI technique that enables non-invasive quantification of brain tissue microstructure through parameters including the Neurite Density Index (NDI) and Orientation Dispersion Index (ODI) (9). These metrics provide more refined biological markers for investigating HE-related cerebral microstructural abnormalities. Recent studies have begun applying NODDI to explore microstructural changes in HE, though most have focused primarily on white matter alterations (10).

Gray-matter Based Spatial Statistics (GBSS) is a voxel-based analysis method specifically designed to investigate spatial patterns of gray matter microstructure (11). A recent study combining NODDI with GBSS successfully identified cortical microstructural abnormalities in patients with minimal hepatic encephalopathy (12). However, this research did not specifically focus on the GP—a region known to be particularly vulnerable in HE—nor did it systematically examine the correlation between cerebral microstructural parameters and blood biochemical indices.

Given these research gaps, the study aims to: (i) assess brain microstructural alterations in patients with HE using NODDI combined with GBSS methodology; (ii) analyze correlations between the NODDI parameters of the GP in the HE group and blood biochemical indices to provide novel imaging evidence for understanding the pathophysiological mechanisms underlying HE.

The demographic information is shown in Table 1.

In this study, we observed widespread neurite microstructural alterations in patients with HE, quantified using NODDI. HE patients showed significantly decreased NDI across multiple gray matter regions including frontal, temporal, parietal, occipital cortices as well as subcortical structures (thalamus, hippocampus, insula), and increased ODI primarily in cerebellar regions. Furthermore, correlations between GP and blood biochemical indices were also identified. These results provide novel biophysical insights into the neuropathology of HE, suggesting a complex interplay of neuroinflammation, edema, and metabolic toxicity.

The most prominent finding was the extensive reduction of NDI in the HE group, with the largest clusters located in the right medial superior frontal gyrus, right insula, and bilateral cingulate gyri. NDI serves as a sensitive marker for the intracellular volume fraction of axons and dendrites. The reduction of NDI in these regions likely reflects a combination of neurite atrophy and the expansion of the extracellular space due to low-grade cerebral edema (13). These regions, which play a central role in executive function, attention, and emotional regulation, may reflect the underlying neuronal damage and dysfunction in HE, consistent with cognitive decline observed in these patients (14, 15).

Interestingly, a recent GBSS-NODDI study (9) similarly identified reduced NDI patterns in patients with minimal hepatic encephalopathy (MHE), demonstrating significant NDI decreases specifically in the left insula and left middle frontal gyrus. These regions are involved in higher-order cognitive and multisensory integration. The consistency of these findings across both HE and MHE populations confirms that microstructural degeneration, detectable by NODDI, is a central feature of HE-related brain injury, beginning even at the clinically covert (MHE) stage. It is noteworthy that the spatial distribution of the most significant alterations differed between the studies, which may reflect the progression of neuropathology. In our HE group, the largest effect sizes were located in the right medial superior frontal gyrus, right insula, and bilateral cingulate gyrus. In contrast, the changes reported in MHE exhibited a left-hemisphere predominance. This pattern suggests a potential trajectory: early, subtle damage may preferentially affect or be more easily detected in left-hemispheric hubs of the executive and salience networks. As liver dysfunction progresses and clinical symptoms emerge, the damage becomes more bilateral and widespread, involving key midline structures, such as the cingulate cortex, and critical right hemisphere regions associated with attention and self-regulation, such as the medial prefrontal cortex.

The HE group also exhibited lower NDI in the precuneus compared to the HC group. The involvement of the medial prefrontal cortex and the precuneus highlights the vulnerability of the Default Mode Network (DMN). Consistent with our findings, previous resting-state functional MRI studies (16, 17) have confirmed the presence of dysfunction in the DMN in patients with HE, characterized by decreased amplitude of low-frequency fluctuation (ALFF) and attenuated functional connectivity. Disruption of the DMN is consistently associated with the cognitive deficits observed in HE, such as impaired attention and self-referential processing (18, 19). Our results extend these findings by pinpointing the biophysical basis of this dysfunction: a reduction in the model-derived neurite density index or integrity. Furthermore, the significant involvement of the insula (part of the Salience Network) supports the “network switching” hypothesis, where structural damage prevents the brain from effectively toggling between central executive and default mode states, leading to the behavioral rigidity seen in patients (20).

We observed a distinct pattern in the cerebellum, characterized by increased ODI in the posterior cerebellum and vermis. While reduced NDI indicates tissue loss, an increased ODI reflects a more disorganized or complex neurite configuration (9). In the context of HE, this increased complexity might represent a maladaptive response or “disorganized branching” in the face of ongoing neurotoxicity. The posterior cerebellum is increasingly recognized for its role in non-motor cognitive processes, including emotional regulation and attention (21, 22). The coexistence of reduced NDI and increased ODI suggests that the cerebellum undergoes active, albeit pathological, remodeling as part of a whole-brain network failure.

Bilirubin is a marker of liver failure severity; its correlation with GP NDI might reflect the synergistic effect of metabolic toxins on astrocyte swelling. An exploratory analysis of the GP was conducted due to its known vulnerability in HE. Although no significant group-level microstructural changes were detected in the GP, the correlation patterns suggest the dissociation between NDI and ODI alterations within the GP in relation to blood biochemical markers. This supports the notion that HE involves widespread pathophysiological mechanisms, including metabolic, inflammatory, and vascular derangements, which in turn affect neuronal microstructure (23). We observed that the NDI of the right GP was positively correlated with markers of liver dysfunction (indirect bilirubin and INR). Typically, neurodegeneration involves a reduction in neurite density. However, the GP is the preferential site for manganese (Mn) deposition and ammonium accumulation in patients with chronic liver disease, often manifesting as T1-weighted hyperintensity (24, 25). This accumulation often leads to astrocyte hypertrophy, which increases the apparent intracellular volume fraction (23, 26). Thus, the “pseudo-increase” in NDI likely reflects toxic cellular swelling rather than healthy neurite growth (27).

Conversely, the left GP ODI showed a positive correlation with hemoglobin concentration. Anemia is a common systemic complication of cirrhosis that exacerbates cerebral hypoxia (28). Our finding suggests that lower hemoglobin levels (more severe anemia) are associated with reduced ODI, representing a simplification of dendritic arborization or loss of synaptic complexity (29, 30). This allows us to propose a hypothesis a “double-hit” mechanism: toxic metabolic accumulation (bilirubin/manganese) drives cellular swelling and increases NDI, while systemic factors like anemia and hypoxia lead to a break-down of structural complexity (reduced ODI) (31) Care should be taken that these correlations should be viewed as exploratory associations; causal inference is not possible with the current cross-sectional design.

Furthermore, our findings should be interpreted within the broader context of metabolic alterations in HE. Previous MR spectroscopy (MRS) studies have consistently demonstrated a characteristic metabolic pattern in HE patients, marked by increased glutamine/glutamate (Glx) and decreased myo-inositol (mI) levels (32, 33). These shifts reflect osmotic abnormalities and astrocyte dysfunction: ammonia-induced glutamine accumulation in astrocytes leads to an osmotic gradient that draws water into the cells, resulting in low-grade brain edema (32, 34). These metabolic shifts provide a critical physiological backdrop for the microstructural changes detected by NODDI in our study, as osmotic stress and subsequent cellular swelling likely contribute to the increased NDI observed in the globus pallidus.

Several limitations in this study should be considered. First, the clinical heterogeneity of decompensated cirrhosis patients, including varying etiologies and liver dysfunction severity, may have introduced confounding factors. Future studies with larger, stratified cohorts are needed to enhance data homogeneity. Second, the sample size was relatively small for an exploratory neuroimaging analysis, which may limit the statistical power to detect subtle cortical alterations. Third, the cross-sectional design precludes the establishment of causal relationships between systemic biochemical changes and microstructural decay. Longitudinal research is essential to evaluate the efficacy of NODDI metrics in predicting disease progression. Fourth, blood ammonia levels were not available for the majority of HE patients due to incomplete clinical records, precluding correlation analyses between this key pathophysiological marker and NODDI metrics. Given the central role of hyperammonemia in HE pathogenesis, future studies should systematically collect ammonia data to clarify its relationship with microstructural alterations. Fifth, the 3.0 T MRI protocol may limit spatial resolution for smaller subcortical structures. Future studies using high-resolution, multi-shell diffusion protocols could further refine these findings. Despite these limitations, the present results provide meaningful insights into HE-related microstructural alterations.

This study employed the NODDI to characterize gray matter microstructural pathology in HE, revealing widespread microstructural alterations consistent with reduced neurite density index and cerebellar disorganization. The observed dissociated correlation patterns in the GP may be interpreted within the framework of a speculative “double-hit” hypothesis of neurotoxicity, although further longitudinal and mechanistic studies are required to validate this interpretation.

In conclusion, NODDI-derived parameters offer sensitive and biologically specific biomarkers that bridge systemic biochemistry and cerebral microstructure, holding promise for improving the early detection and mechanistic evaluation of HE.