Machine Learning in Classification of Parkinson’s Disease Using Electroencephalogram with Simon’s Conflict

Abstract

Parkinson's disease (PD) is one of the most common health problems globally. It is essential to diagnose healthy people and patients with PD precisely. This study contains 28 patients who tested ON and OFF dopaminergic medication and 28 healthy people of matched age and sex. Their electroencephalography (EEG) signals, which were the most frequently employed approach for diagnosing brain illness, were collected while participating in Simon's Conflict for the classification phase. The proposed method consists of preprocessing, feature extraction, and classification. This research used Independent Component Analysis to separate the EEG signal into several components to remove noise. Discrete wavelet transforms (DWT) decomposed the EEG signal to delta, theta, alpha, beta, and gamma bands to extract important information. This research emphasizes using ICA, DWT, and machine learning algorithms to diagnose healthy people and PD. Research also indicated that removing noise from EEG signals was a crucial step, which helped to improve the result significantly. The classification result using linear Support Vector Machine gives the highest performance: 84.5% accuracy, 89.3% sensitivity, and 75.0% specificity.