MADNet: EEG-Based Depression Detection Using a Deep Convolution Neural Network Framework with Multi-dimensional Attention

Shuyu Chen, Yangzuyi Yu, Jiahui Pan

Abstract: Major depressive disorder (MDD) is a common and serious mental health problem that has received increasing attention from both researchers and clinicians. Electroencephalography (EEG)-based automatic diagnosis of MDD has been explored in previous studies, but feature extraction remains an area for improvement. We propose a novel deep learning approach for the automated detection of MDD. For feature extraction, a multi-dimensional attention mechanism is used to extract features from a simple pre-processed EEG signal to enhance spatio-temporal feature learning. We randomly divided EEG data from 26 MDD patients and 26 healthy controls into training and testing sets for network training and testing, respectively. Our method achieved the highest accuracy of 95.14% in subject-based data classification experiments compared to other methods. As a result of its simplicity and convenience, the method proposed in this study has the potential to be utilized as a tool for assisting in the daily diagnosis of depression.