Published Paper Details:

Background: Depression has become the number one mental disorder worldwide. It has been shown that subjects with depression exhibit different spatial responses to neurophysiological cues from healthy controls when exposed to positive and negative stimuli. Methods: We propose an efficient EEG-based detection method for classification of depression using spatial information. Thirty participants, including 16 depressed patients and 14 healthy controls, were given a face-to-face task that included positive and negative emotional facial expressions. Differential entropy and genetic algorithms are used for feature extraction and selection, and support vector machines are used for classification. A task-dependent common spatial pattern (TCSP) has been proposed to enhance the spatial disparity before feature extraction. Results and Discussion: We obtained TCSP positive and negative stimulus classification scores of 84% and 85.7%, respectively, which were statistically significantly higher than those obtained without TCSP of 81.7% and 83.2%, respectively (p < 0.05). We also evaluated the classification performance using individual frequency bands and found that the contribution of the gamma frequency band dominates. Furthermore, we evaluate various classifiers, including k-nearest neighbors and logistic regression, which show similar trends in classification improvement by using TCSP.