Diagnosis of Parkinson’s disease using higher order statistical analysis of alpha and beta rhythms

Parkinson’s disease (PD) is one of the most common neurodegenerative diseases and is generally associated with its signature symptoms of rest tremor, muscle rigidity and bradykinesia. Currently, PD is diagnosed by neurologists who focus on consider multiple factors to make their decision. Biomarkers such as electroencephalography (EEG) signals can be used for the classification of PD from healthy control (HC). These methods offer an objective approach and can act as an aid for neurologists in the PD diagnosis process. In this study, we introduce new higher order statistical (HOS) features of EEG signals derived from the alpha and beta rhythms and use them for classification of PD from HC using ensemble learning. This machine learning approach helps to improve the result of classification by combining multiple models and produces a better predictive performance compared to a single classification model. Our approach is able to achieve an average sensitivity of 99.28% with 99.10% specificity using the Bagged trees ensemble classifier. These results compared to previous studies conducted in this field demonstrate the importance of HOS and different rhythm features in background EEG analysis along with the superiority of ensemble classifiers for these types of applications compared to other machine learning and deep learning methods.