Evaluation of cortical lateralization for identifying Parkinson's disease patients using electroencephalographic signals and machine learning

Parkinson's disease (PD) is one of the most common neurodegenerative diseases in the world and occurs due to the death of dopaminergic neurons, responsible for controlling body movements. PD brings symptoms of motor and non-motor origin, but the diagnosis depends on the appearance of motor symptoms, which occur when the disease is already in an advanced stage. Therefore, several studies have been developed to obtain biomarkers that allow the diagnosis of PD through artificial intelligence algorithms. In this context, electroencephalogram (EEG) signals can be used. In addition, it is known that PD patients have one side of the body more affected by motor symptoms, which means that the opposite hemisphere of the brain is more affected by the disease. Thus, this project aimed to investigate the use of cortical lateralization in the development of a system to aid the diagnosis of Parkinson's disease using EEG signals. A database of EEG signals collected while participants, with and without Parkinson's, remained at rest or performed the Oddball Paradigm task was used. Statistical and spectral analysis techniques with separation of alpha, beta, and gamma frequencies were used to extract features from the signals, and the performance of these features was investigated using machine learning tools. Accuracies of over 70% were obtained and it was observed that cortical lateralization improves model performance.