چكيده به لاتين
Today, depression is a common mental illness recognized as a social problem worldwide, associated with low mood and dysfunction. Therefore, accurate and early identification of depression is one of the current challenges. Recently, the automatic diagnosis of depression by electroencephalogram (EEG) signals, a non-invasive, cheap, robust, and high-time resolution tool, has attracted much attention. This research extracted linear features in the whole frequency band, non-linear features, statistical features in the whole frequency band and five sub-bands, and features based on the cross-frequency coupling to classify depressed patients and healthy groups. Twenty-two patients with major depresive disorder and 15 healthy individuals participated in this research. After feature extraction, features with statistically significant were selected by Wilcoxon and Kruskal-Wallis statistical tests and classified by support vector machine, k-nearest neighbor, linear discriminant analysis, decision tree, and logistic regression. Statistical analysis and classification results of linear, non-linear, and statistical features showed the most statistically significant differences between the two groups in the right occipital regions and the left temporal region (T5) in the beta band. Statistical analysis and classification results in features based on cross-frequency coupling showed better performance in frequency-frequency coupling and phase-phase coupling. Frequency-frequency and phase-phase coupling acted as a potential biomarkers, and all ten features selected from frequency-frequency coupling individually achieved 100% accuracy. We also found that the combination of the beta band and gamma band increases the classification accuracy. In addition, features based on cross-frequency coupling had better performance than other measures.
