Classification of schizophrenia using general linear model and support vector machine via fNIRS

Schizophrenia is a type of serious mental illness. In clinical practice, it is still a challenging problem to identify schizophrenia-related brain patterns due to the lack of objective physiological data support and a unified data analysis method, physicians can only use the subjective experience to distinguish schizophrenia patients and healthy people, which may easily lead to misdiagnosis. In this study, we designed an optimized data-preprocessing method accompanied with techniques of general linear model feature extraction, independent sample t-test feature selection and support vector machine to identify a set of robust fNIRS pattern features as a biomarker to discriminate schizophrenia patients and healthy people. Experimental results demonstrated that the proposed combination way of data preprocessing, feature extraction, feature selection and support vector machine classification can effectively identify schizophrenia patients and the healthy people with a leave-one-out-cross-validation classification accuracy of 89.5%.