Development of a clinical prediction model for inflammatory biomarkers and enlarged basal ganglia perivascular spaces using SHAP analysis: feature selection and model interpretation.

Background: Enlarged perivascular spaces (EPVS) may lead to dysfunction of the cerebral lymphatic system, which may be associated with cerebrovascular diseases, cognitive dysfunction, and other neurological diseases. However, the association between cognitive function and systemic inflammation has not been systematically elucidated. This study aimed to develop a predictive model integrating the Montreal Cognitive Assessment (MoCA) and complete blood count-derived inflammatory markers to analyze the relationship between multidimensional indicators and BG-EPVS burden. Methods: We consecutively enrolled patients with cerebral small vessel disease (CSVD) admitted to the Department of Neurology, First Affiliated Hospital of Baotou Medical College, between 2023 and 2024. BG-EPVS severity was evaluated using MRI, and statistical analyses were conducted on clinical variables. Univariate and multivariate logistic regression analyses were conducted to identify independent predictors of BG-EPVS severity. Model performance and clinical utility were evaluated using receiver operating characteristic curves (ROC-AUC), calibration plots, decision curve analysis (DCA), and clinical impact curves (CIC). Model interpretability was assessed using SHapley Additive exPlanations (SHAP). Results: Multivariate logistic regression identified MoCA score, age, hypertension, SIRI and education independent predictors of BG-EPVS burden. Conclusions: These findings demonstrate that age, hypertension and SIRI were positively correlated with high BG-EPVS burden, while MoCA score and education duration were negatively correlated. The integrated model combining MoCA and inflammatory biomarkers accurately predicts BG-EPVS burden, demonstrating their clinical value in early disease screening and dynamic monitoring.