Interpretable Machine Learning for Cognitive Impairment Prediction in Parkinson's Disease: A Multicenter Validation Study with SHAP Analysis

Ziyuan Wang^1,2*Junqiang Yan^1,2*

• ¹Henan University of Science and Technology, Luoyang, China
• ²The First Affiliated Hospital of Henan University of Science and Technology, Luoyang, China

This study developed an interpretable machine learning framework to detect Parkinson's disease-related cognitive impairment (PD-CI) using routine clinical data. We analyzed 1,279 participants from the Parkinson's Progression Markers Initiative (PPMI) discovery cohort and 197 patients from an independent validation cohort, with PD-CI defined by MoCA ≤26 and UPDRS-I ≥1. Twenty-one clinical features—including hematological parameters, metabolic markers, and demographics—were preprocessed using synthetic minority over-sampling. Four machine learning models were trained via nested 5-fold cross-validation, with the Random Forest algorithm achieving superior performance (AUC=0.83), outperforming CatBoost (0.82), XGBoost (0.79), and neural networks (0.66). External validation preserved 71.57% accuracy. SHAP interpretability analysis identified age, neutrophil-to-lymphocyte ratio (NLR), and serum uric acid as critical predictors, revealing synergistic risk effects between elevated inflammation markers and reduced antioxidant levels. The framework demonstrated diagnostic accuracy comparable to advanced neuroimaging while highlighting neuroinflammation and oxidative stress as mechanistic drivers. Multicenter validation confirmed robustness across ethnic populations, supporting its utility in clinical monitoring.