Schizophrenia Detection Using Convolutional Neural Network

Abstract

Paper deals with the recognition of cognitive impairment schizophrenia based on the eye movements of two groups of individuals - healthy and diagnosed. Eye movements tracking is an effective method for examining the relationship between a subject's behavior and cognitive functions. Since there is still not common usage of automatic diagnostic tools in the field of medical diagnosis, specifically psychiatry, our proposed approach presents method which could be helpful as preclinical diagnostic tool. In our method we are using Convolutional Neural Network (CNN) for classification of the saliency maps, gained from gaze raw data, measured when subjects were exposed to Rorschach inkblot test (ROR). Clinical sample of tested subjects consists of 24 healthy and 24 diagnosed individuals. The best average accuracy of classification is 74.44%.