Automated Depression Diagnosis in MDD (Major Depressive Disorder) Patients Using EEG Signal

The detection of major depression is very critical process. The identification and treatment of depression at right time is very essential for well-being of person along with family and society. The acquisition of signal from patient is complex and time-consuming process. So, Multi-modal Open Dataset for Mental-disorder Analysis (MODMA) is considered to accomplish present research work. Different modalities are available for identification of mental stress but here in this work, Electroencephalography (EEG) technique is chosen due to its painless and low-cost features. All the relevant 10 features (linear and non-linear features) are calculated from the dataset of 10 subjects (5 MDD and 5 Healthy Control (HC)) using EEG LAB toolbox in MATLAB R2020b software. The array of matrix of various features is formed for all subjects (5 MDD and 5 HC). All 128-channel EEG data features calculated in more effective way. The classification process is accomplished using 5 Classifiers named Linear SVM (Support Vector Machine), Fine Tree, LR (Logistic Regression), Kernel Naïve Bayes and Fine KNN (K-Nearest Neighbor) for better accuracy. The highest average correct classification rate for Fine Tree classifier for 5 features and 10 features is found to be 99.52% and 99.68%, respectively. These results are compared with other classifiers to reach final conclusion of finding best modality to differentiate MDD patient from healthy one. In this way, we find out best classifier for detection of depression using MODMA dataset of 128-channel EEG signal.