Depression Detection Using Spatial Images of Multichannel EEG Data

Abstract

Depression or Major Depressive Disorder (MDD), is a common mental disorder and globally, more than 300 million people of all ages suffer from it. Depression is a leading cause of disability worldwide and is a major contributor to the overall global burden of disease. An approach for detecting depression from multi-channel EEG time-series data has been proposed. A computer-aided Machine learning approach: Convolutional Neural Network (CNN), a deep learning method is used in this work. The publicly available EEG dataset is gathered from a study of depression in which the brain activity of currently depressed and normal individuals is recorded using 67 channels in separate files. Firstly, the EEG activities are transformed into a sequence of topology-preserving multispectral images. Next, a deep recurrent-convolutional neural network is trained to learn the robust representations from the sequence of images. This model is beneficial as it preserves the spatial, spectral, and temporal structure of EEG data. The results depict that the proposed deep learning method could achieve better classification performance on large amounts of data.