An AI-Based Predictive Framework For Parkinson’s Disease Diagnosis

Abstract

Parkinson disease (PD) is a progressive neurodegenerative illness that affects motor and non motor functions greatly and thus must be diagnosed early in order to clinical care. Conventional methods of diagnosis are based more on clinical observation thus leading to sluggish or unreliable detection. In this paper, we outline an AI predictive framework to diagnose Parkinson disease that uses the interaction of sophisticated machine learning algorithms and explainable AI methods and strategies to increase the level of diagnostic accuracy and interpretability. The framework relies on the UCI Parkinson data which takes the form of biomedical voice measurements and involves the data preprocessing, feature selection, and dimensionality reduction to obtain the most significant biomarkers. Various machine learning algorithms, such as the Random Forest, Support Vector machine, and Gradient Boosting are trained and tested with the typical performance measures, including the accuracy, precision, recall, F1 score and ROC AUC. The explainable artificial intelligence tools like SHAP (SHapley Additive explanations) can inform about the significance of features and the decisions made by the model, which orients to transparent use in clinical settings. Through experimental findings, the proposed framework has high predictive accuracy despite giving interpretable results and is therefore useful in identifying early stage Parkinson-related disease. The analysis shows the possibilities of AI driven solutions as valid decision support tools, which will allow providing timely intervention and better patient care.