MDD Detection based on Empirical Mode Decomposition and Effective Brain Network Analysis

Depression is a psychological disorder recognized by symptoms such as a feeling of hopelessness and sadness which can lead to suicide if not diagnosed and treated on time. Electroencephalography (EEG) is a powerful tool due to its applicability in depression detection and it is more difficult to find appropriate biomarkers from all channels. In this paper, to detect depression, a new method is introduced based on empirical mode decomposition (EMD) and effective connectivity (EC) analysis. The performance of the proposed algorithm is verified on a publicly available depression database. The results show that the first intrinsic mode function (IMF) decomposed by EMD has significant effect on EEGs of patients with major depressive disorder (MDD) compared with healthy control. At the next step, the connectivity matrices of all subjects obtained from Granger causality method are used to extract different features. Then, the connectivity measures are fed to the convolutional neural network (CNN) classifier. The proposed method as compared with state-of-the-art depression detection approaches from literature yields the best performance in terms of accuracy, ۹۹.۸۹%. this means that the proposed technique can be used in healthcaredevices to identify the MDD patients for early intervention.

Keywords:

Major Depressive Disorder , Electroencephalography , Empirical Mode Decomposition , Effective Connectivity , Convolutional Neural Network.