Diagnosis of Major Depressive Disorder from EEG Signals Using Empirical Wavelet Transform and Machine Learning Methods

2024 Medical Technologies Congress (TIPTEKNO), Muğla, Turkey, 10 - 12 October 2024, pp.1-4, (Full Text)

Publication Type: Conference Paper / Full Text
Doi Number: 10.1109/tiptekno63488.2024.10755410
City: Muğla
Country: Turkey
Page Numbers: pp.1-4
Istanbul University-Cerrahpasa Affiliated: Yes

Abstract

In this study, Empirical Wavelet Transform (EWT), a time-frequency analysis method, was applied to EEG signals to diagnose Major Depressive Disorder (MDD). The EEG signals in the Fp1, Fpz, and Fp2 channels of MDD patients and healthy individuals resting in the MODMA dataset were analyzed in delta and theta frequency bands. By applying EWT to EEG signals, three channels of EEG signals were decomposed into 6 sub-signals. A total of 126 features were obtained by extracting 21 features from each of these sub-signals. The obtained features were classified using machine learning (ML) algorithms (Decision Trees, Support Vector Machine, k-nearest Neighbor, Ensemble Bagged Trees, Wide Neural Network). According to the results of the analyses, it was observed that the Theta band was more effective in distinguishing the MDD patients and healthy individuals. Additionally, the Ensemble Bagged Trees classifier provided the highest accuracy rates. This algorithm provided 98.88% accuracy in the Theta band, but its performance was lower in the Delta band like the other classifiers. This study emphasizes the promising potential of EWT based features in MDD diagnosis and their potential to improve diagnostic processes. The accuracy results demonstrate the usability of EWT in MDD diagnosis and the effectiveness of ML techniques in this process.