Depression Classification and Recognition by Graph-Based Features of EEG Signals

Major depressive disorder (MDD) is one of the main subjects in world health so its diagnosis is important for researchers. Electroencephalography (EEG) is one of the effective tools in brain psychological disorders diagnosis in which any change in brain function is reflected in the signals. By analysing the EEG signals, some disorders like MDD can be recognised. In this paper, EEG signals are firstly mapped to four different visibility graphs and several features are extracted from each graph. Then feature numbers are reduced by principal component analysis (PCA) and depressive and normal classification is done by support vector machine (SVM). In this paper, classification results by combining all four graph features are compared with each graph features individually, and the results show that by the combining features, lower classification error and better accuracy is achieved. The classification accuracy of depression classification by mixed features is 100% which means the proposed method can classify all of them correctly.