Diagnosis of Parkinson's disease in imagining right and left hand movements from entropy information of wavelet EEG signal subbands a nd combination of classical classifications

Background: Using of artificial intelligence to diagnose diseases, including Parkinson's disease, are increasing. It is the second most common neurodegenerative disease Therefore, designing an accurate detection system to diagnose the disease is critical. Methods: In this paper, delta, theta, alpha and beta frequency bands are extracted from ۱۹-channel EEG signals. The signals were recorded at a frequency of ۲۵۶ Hz while imaging motion from these two groups. In the next step, two nonlinear samples of sample entropy and Shannon were extracted from ten sections of each EEG channel After normalization with conventional methods such as zscore, the dimensions of the feature space were reduced to ۵ basic components using the principal component analysis (PCA) method and the classification had been done in two scenarios. In the single-classification scenario, three methods of support vector machine (SVM), multilayer perceptron (MLP) and K-nearest neighbor (KNN) are used separately.Results: The highest accuracy of Parkinson's disease diagnosis was obtained from EEG signals following the combination of categories by the majority voting method of ۹۳/۴۵ ± ۲/۳. Then, thehighest accuracy of Parkinson's disease diagnosis was obtained from EEG signals using SVM classification method with a triangular kernel of ۸۷/۶۰ ± ۲/۶.Conclusions: The results of the certificate show the correct method in diagnosing Parkinson's disease from EEG signals of sick and healthy people while imagining movement.

کلیدواژه ها:

Parkinson's disease ، electroencephalogram (EEG) signal ، sample entropy ، Shannon entropy ، support vector machine (SVM)