Comparative Analysis of Brain Waves for EEG-Based Depression Detection in the Prefrontal Cortex Lobe using LSTM

Global anxiety and depression have become 25% more prevalent, with teenagers and women being the most affected. Approximately 280 million people suffer from depression. Doctors and psychologists are able to diagnose depressive disorders through counselling sessions and ask relevant questions to the subject, despite being vulnerable to mistakes due to the examiner's lack of experience. Therefore, automated depression detection development is necessary to validate doctor and psychiatrist assessment. Electroencephalography (EEG) is considered to be a popular option for the detection and investigation of various mental disorders. In this study, a comparison and analysis of each existing brain wave is carried out, namely Alpha (8-12Hz), Beta (13-30Hz), Theta (4–8 Hz), Delta (0.5-4 Hz) and Gamma (30-50Hz). From each wave, an accuracy testing is carried out for three groups of features: linear features, nonlinear features, and a combination of linear and nonlinear features. The given results demonstrate that the combination of linear and nonlinear data consistently yields the highest accuracy outcomes across all waves. Also, the combination of theta waves and linear nonlinear features contributed the highest accuracy (84%) using LSTM as the classifier.