Classification of Parkinson's Disease Using Motor and Non-Motor Biomarkers Through Machine Learning Techniques

Abstract

Parkinson's disease (PD) is the second most neurodegenerative disease in the United States of America after Alzheimer's disease. The Parkinson's disease patients and scans without evidence for dopaminergic deficit (SWEDD) patients will share the same symptoms, and It's hard to differentiate the PD, SWEDD patients, and healthy controls in the progression of PD. In this research, we classified PD patients, SWEDD patients, and healthy controls by considering motor and non-motor biomarkers, namely MDS-UPDRS part 1, SCOPA score, and QUIP-RS from the PPMI database by using supervised and unsupervised machine learning algorithms, namely Knn, logistic regression, XGBooting, naive Bayes, Decision tree, Random Forest, Support vector machine, multilayer perceptron , and K-means clustering, respectively. Random Forest scored 0.98 percent accuracy among all these algorithms and can identify and differentiate PD, SWEDD, and Healthy controls patients by motor and non-motor biomarkers.