An Intelligent CDS (Clinical Decision Support) Framework using Machine Learning Algorithms for Parkinson Disease Detection

Parkinson’s disease is a progressive neurodegenerative disorder characterized by the gradual onset of symptoms, complicating early diagnosis. Traditionally, neurologists diagnose Parkinson’s through patient medical history reviews and repeated scans, while body movement analysts focus on evaluating physical movements. Recent studies suggest that speech alterations can serve as quantifiable markers for early Parkinson’s detection. This paper introduces a novel and robust framework that leverages the Multiple Feature Evaluation Approach (MFEA), the Synthetic Minority Oversampling Technique (SMOTE), and Neural Networks for the early detection of Parkinson’s disease. A new algorithm, termed Multi-Agent MFEA, is proposed within this framework, which is further developed into an intelligent system called iCDS (Intelligent Clinical Decision Support). The effectiveness of this system is validated using a speech dataset from the UCI Machine Learning Repository. Comparative analysis shows that the iCDS framework, equipped with MFEA, outperforms traditional feature selection methods (such as Mutual Information Gain and Recursive Feature Elimination) and classifiers (including Naive Bayes, KNN, and SVM) in multiple aspects. The study’s findings have the potential to aid both patients and medical professionals in making timely decisions and improving medical support. The development of Clinical Decision Support (CDS) systems has made early detection of Parkinson’s disease more feasible, and this research substantiates that advancement.