Bi-objective approach for computer-aided diagnosis of schizophrenia patients using fMRI data

Abstract

Computer-aided diagnosis (CAD) of schizophrenia based on the analysis of brain images, captured using functional Magnetic Resonance Imaging (fMRI) technique, is an active area of research. The main problem lies in the identification of brain regions that contribute to differentiating between a healthy subject and a schizophrenia affected subject. The problem becomes complex due to the high dimensionality of the fMRI data on the one hand and the availability of data for only a small number of subjects on the other hand. In this paper, we propose a three-stage evolutionary based framework for feature selection. It comprises application of general linear model, followed by statistical hypothesis testing, and finally application of Non-dominated Sorting Genetic Algorithm (NSGA-II) to arrive at a small set of about fifty features. Experiments show that the feature set generated by the proposed approach yields accuracy as high as 99.5% in classifying fMRI dataset of healthy and schizophrenia subjects, and can identify the relevant brain regions that are affected in schizophrenia.