Detecting Parkinson’s Disease from Electroencephalogram Signals: An Explainable Machine Learning Approach

Abstract

Parkinson’s disease (PD) is the second most common neurological disorder. It is characterised by stiffness, rigidity, tremor, freezing gait and postural instability. PD is monitored clinically by expert neurologists by visually inspecting upper and lower limb movements, speech, gait and facial expressions. This is time-consuming, error-prone and requires an expert neurologist to perform these manual inspections. The electroencephalogram (EEG) is a non-invasive method of monitoring brain activity. This work proposes an EEG-based automated PD monitoring technique. PD was identified using explainable machine learning classifiers based on 31 features extracted from EEG signals. To distinguish PD from healthy controls, the support vector machine classifier with a polynomial kernel achieves 87.10% accuracy, 93.33% sensitivity and 81.25% specificity.