Metabolic dysfunction-associated fatty liver disease exacerbates hematoma expansion in intracerebral hemorrhage: an explainable machine learning approach.

BACKGROUND: Although hematoma expansion (HE) severely deteriorates outcomes in hypertensive intracerebral hemorrhage (HICH), predicting HE in patients comorbid with metabolic dysfunction-associated fatty liver disease (MAFLD) lacks specific machine learning tools. This study aimed to bridge this gap by exploring the interactive pathophysiological associations. METHODS: Data from 529 HICH-MAFLD patients were used to construct five predictive algorithms. The best-performing model was subjected to SHapley Additive exPlanations (SHAP) to map variable interactions. RESULTS: The XGBoost model achieved the best discrimination (AUC = 0.911). SHAP analysis revealed non-linear, synergistic statistical associations between the "liver-brain axis" components. Specifically, severe hepatic fibrosis (FIB-4 index) and systemic inflammation (hs-CRP) appeared to exponentially amplify the HE risk associated with local hemodynamic parameters (baseline hematoma volume and systolic blood pressure). CONCLUSION: The SHAP-integrated XGBoost algorithm suggests a robust predictive framework for early HE. The statistical synergy unmasked between metabolic inflammation and local hemodynamics highlights the potential clinical significance of the "liver-brain axis" in guiding individualized risk stratification.