Depression Detection in Working Environment using 2D CSM and CNN with EEG Signals

Treatment of depression, which is a common illness around the world, remains challenging without early detection. The electroencephalogram (EEG), a non-invasive form of functional imaging, has been widely used to discover the key biomarkers for understanding the disorder. Recent advances have been made in computerized depression diagnosis that uses machine learning algorithms. Identifying in advance which treatment agent is most effective for different psychiatric disorders remains a challenge. We proposed a deep Convolution Neural Network (CNN) architecture to detect depression using 2D Cyclic Spectrum Map (CSM) images. The 2D Cyclic Spectrum Map (CSM) is applied to EEG signals to create spectral images for Depressive Disorder (DD) patients and healthy people. The 2D cyclic spectrum Map images are generated from EEG signals, thereby spectral images which will be classified through CNN into Normal, Anxiety, High Blood Pressure, and Cardiovascular disease. Experimental results show that the CNN-2D CSM model attains 99.5% accuracy and 99.3% sensitivity than the existing (CNN-LSTM) method (accuracy - 99.12% and sensitivity - 99.11%) in fast detecting the EEG signals therefore, it can be utilized in psychiatry wards of the hospitals.