Machine Learning and Explainable AI for Parkinson’s Disease Prediction: A Systematic Review

Abstract

Parkinson's disease is a movement disorder within the nervous system that impacts millions of people across the world. The standard diagnostic methods usually miss early subtle signs of disease which has driven research into Machine Learning (ML) and Explainable Artificial Intelligence (XAI) to develop better predictive models. Following PRISMA guidelines we analyzed 13 studies found in IEEE Xplore, PubMed and ACM concerning different ML methodologies for Parkinson’s disease prediction. Deep learning models using vocal and motor data achieve good accuracy but require more clinical trust and transparency due to their opaque "black-box" nature. SHAP and LIME act as XAI solutions that address transparency issues in model predictions by delivering understandable information regarding model outputs to all users. Current solutions show progress. However, there are multiple complications, including limited and unbalanced datasets alongside accuracy-explainability trade-offs which underline the need for extensive datasets, multidisciplinary teamwork and practical validation.