Parkinson's Disease Diagnosis from Patients Speech Analysis

Parkinson's disease is a neurological disorder mostly observed in the population over the age of 60 years. Over 8.5 million people world-wide are affected by it. Early diagnosis of the disease helps take preventive and curative measures to enable the affected individuals to lead a normal life. Quality of speech is one of the key indicators for identification of the disease.In this study, we investigate the effectiveness of multiple machine learning algorithms on various speech related attributes to determine the existence of this disease. Additionally, we identify the crucial features which impact the model's functioning. The study employs Explainable AI (XAI) to investigate the influence of voice features on the research outcomes. The experimental findings showcase a diagnostic accuracy of 95% using Random Forest & XGBoost classifiers. We, using SHAP, identify that 6 features out of a set of 24 provide the best efficiency of diagnosis.