An Explainable Approach for Early Diagnosis of Parkinson's Disease Exploring Motor Symptoms

Parkinson's disease is a progressive neurodegen-erative disease that affects the nervous system causing severe cognitive, motor, and behavioral dysfunctions. It is mainly caused by a progressive loss of neurons in the substantia nigra in the basal ganglia and this leads to abnormalities in the control of movements. The symptoms progress slowly over the years and for this reason, the diagnosis of Parkinson's disease is very complex, so it is important to support medical personnel in identifying and predicting the disease with Artificial Intelligence tools and techniques. This study proposes a new approach to predict disease onset and monitor patient progress through Explainable Artificial Intelligence (XAI) that aims to provide clear and understandable explanations for decisions made by machine learning algorithms. Initially, different classification algorithms were evaluated and, by choosing the one with the best performance, the most significant features were identified through the use of SHAP. The model was subsequently re-evaluated with only the selected features, evaluating the use of the XAI for disease identification. The methodology was validated on the Parkinson's Progression Markers Initiative (PPMI), more specifically taking into consideration only the characteristics that concern the motor sphere.