Imagery Signal-Based Deep Learning Method for Prescreening Major Depressive Disorder

Depression is a high-risk mental illness that can lead to suicide. However, for a variety of reasons, such as a negative perception of mental illness, most patients with depressive symptoms are reluctant to go to the hospital and miss appropriate treatment. Therefore, a simple prescreening method that an individual can use to identify depression is needed. Most EEG measurement devices that individuals use have few channels. However, most studies using EEG to diagnose depression have been conducted in a professional multi-channel EEG environment. Therefore, it is difficult for individuals to prescreen depression based on the results of the studies. In this study, we proposed a model that predicts depression by using EEG data measured by a few channels so that it can measure depression using the EEG data measured by an individual. In this study, brain waves measured in two channels were imaged using STFT transform and a spectrogram. The EEG image data was then used in a deep learning model. As a result of the performance evaluation, 75% accuracy was shown for the classification of image depression EEGs and normal image type EEGs. As a result, low channel EEG data for deep learning can be used as an auxiliary tool to proactively diagnose depressed patients.