Classification of Schizophrenia by Seed-based Functional Connectivity using Prefronto-Temporal Functional Near Infrared Spectroscopy

Background Schizophrenia is one of the most serious mental disorders. Currently, the diagnosis of schizophrenia mainly relies on scales and doctors' experience. Recently, functional near infrared spectroscopy (fNIRS) has been used to distinguish schizophrenia from other mental disorders. The conventional classification methods utilized time-course features from single or multiple fNIRS channels. New method The fNIRS data were obtained from 52 channels covering the frontotemporal cortices in 200 patients with schizophrenia and 100 healthy subjects during a Chinese verbal fluency task. The channels with significant between-group differences were selected as the seeds. Functional connectivity (FC) was calculated for each seed, and FCs with significant between-group differences were selected as the features for classification. Results The proposed method reduced the number of channels to 26 while achieving overall classification accuracy, sensitivity and specificity values as high as 89.67%, 93.00% and 86.00%, respectively, outperforming most of the reported results. The superior performance was attributed to the cross-scale neurological changes related to schizophrenia, which were employed by the classification method. In addition, the method provided multiple classification criteria with similar accuracy, consequently increasing the flexibility and reliability of the results. Comparison with existing methods This is the first fNIRS study to classify schizophrenia based on FCs. This method integrated information from regional modulation, segregation and integration. The classification performance outperformed most of the classification methods described in previous studies. Conclusions Our findings suggest a reliable method with a high level of accuracy and a low level of instrumental complexity to identify patients with schizophrenia.