Abstract

Schizophrenia is a psychiatric disorder in which the individual’s sense of reality is altered. Currently there are no tests that can absolutely diagnose the disorder. Our aim is to develop a computer aided diagnostic(CAD) tool by tapping the EEG signals from the scalp of the head and extracting the P300 event related potentials(ERPs). This CAD tool can be used to objectively aid in the diagnosis and distinguish between Healthy Controls (HC) and subjects of Schizophrenia (Sz). The proposed methodology consists of four different stages: EEG preprocessing, feature extraction, feature selection and classification. EEG data of 8 schizophrenic and 8 normal subjects were considered for an auditory odd ball task and visual odd ball task. Features are extracted and are analyzed in both time and frequency domain. These features have been subjected to a feature selection algorithm called principal component analysis and the selected features will be used to train three classifiers. The accuracy of these three classifiers will be compared which include linear discriminant analysis, support vector machines and multi-layer Perceptron and the classifier with the highest accuracy is selected to test the data and perform plotting. The accuracy and efficiency of multi-layer Perceptron classifier was found to be better compared to the other methods discussed.