Computer Aided Detection of Major Depressive Disorder (MDD) Using Electroencephalogram Signals

MDD or depression is psychiatric disorder which affects many of people globally. Due to the nonexistence of any available laboratory tests, poor identification of depression is a key contributing factor. In this study, a framework is developed in order detect depression with minimal number of channels to increase the portability as well as with high accuracy. In this study, two dataset has been considered i.e. Public and Private dataset. Different regions of brain as well as feature has been selected to find out the minimal number of channels and features so that higher accuracy is also achieved. According to the study, depression has different impacts on each hemisphere of the temporal and parietal regions of the brain. Depression also has an impact on Detrended Fluctuation Analysis. The study results in a model made up of 4 channels with high accuracy of 91.74% which is portable, faster and cheaper. Thus, model can act as an assistive tool for diagnosis of depression. The study confirms that depression affects the temporal area of the brain since using the same set of features and classifiers in both the public and private datasets and utilising only temporal channel EEG data provided quite high accuracy in detection of MDD. The fact that using features delta, alpha, beta paired asymmetry and DFA from only temporal channel (T8 and T9) provided high accuracy of 89.24%(Public Dataset) and 82.68%(Private Dataset) supports the claim that the temporal lobe of the brain is impacted by depression.