Electrode subset selection to lessen the complexity of brain activity measurement using EEG for depression detection.

Depression is a severe neurological disorder characterized by a loss of interest and may lead to suicide. Electroencephalography (EEG) measurement is a non-invasive tool for neural electrical activities measurement which can be further used for different neurological disorder detection such as depression. The number of EEG electrodes used for measurement directly affects the instrumentation and measurement complexity of the experiment. This paper proposes a fisher score–based method for electrode ranking. This paper selects only those electrodes whose fisher score is greater than the mean of fisher scores of all electrodes. It results in a reduced set of electrodes. A deep learning–based model has been proposed which uses the reduced set of electrodes for depression detection. The performance of the proposed model is evaluated on two benchmark data sets having varying numbers of electrodes. The proposed model significantly reduces the number of electrodes to 68.42% and 60.93% for data sets 1 and 2, respectively. The accuracy of 98.73%, precision of 98.50%, recall of 98.75%, F1 score of 98.62% and AUC of 99.91% are obtained for data set 1 and accuracy of 95.48%, precision of 91.93%, recall of 96.11%, F1 score of 93.97% and AUC of 99.49% are obtained for data set 2. [ABSTRACT FROM AUTHOR]