Explainable AI for Parkinson's disease prediction: A machine learning approach with interpretable models.

Area of Science:

• Neurology
• Artificial Intelligence
• Data Science

Background:

• Parkinson's Disease (PD) presents diagnostic challenges due to its progressive nature and limitations of traditional methods.
• Machine Learning (ML) offers potential for early PD prediction, but interpretability issues hinder clinical use.
• Explainable Artificial Intelligence (XAI) is crucial for bridging the gap between ML models and clinical practice.

Purpose of the Study:

• To develop an interpretable ML model for early and accurate PD prediction.
• To utilize multimodal datasets and XAI techniques for enhanced diagnostic capabilities.
• To improve clinical decision-making and patient care through advanced predictive modeling.

Main Methods:

• Applied five ML algorithms (SVM, KNN, LR, RF, XGBoost) and a stacked ensemble to a Kaggle dataset (n=2105).
• Data included demographics, medical history, lifestyle, clinical symptoms, and cognitive/functional assessments.
• Feature selection (SBE) and interpretation (SHAP, LIME) were performed on the best-performing Random Forest model.

Main Results:

• The Random Forest model with SBE achieved 93% accuracy, precision, recall, and F1-score, with an AUC of 0.97.
• SHAP and LIME identified UPDRS scores, cognitive impairment, functional assessment, and motor symptoms as key predictors.
• XAI techniques successfully enhanced the interpretability of the predictive model.

Conclusions:

• An interpretable Random Forest model effectively predicts Parkinson's Disease.
• Integrating ML and XAI significantly enhances clinical decision-making and diagnostic timing.
• This approach supports personalized patient care and improves management of PD.