Resting-state anticorrelated networks in Schizophrenia

Abstract

Converging evidences from different lines of research suggest abnormalities in functional brain connectivity in schizophrenia. While positively correlated brain networks have been well researched, anticorrelated functional connectivity remains under explored. Hence, in this study we examined (1) the resting-state anticorrelated networks in schizophrenia, and (2) the accuracy of support vector machines (SVMs) in differentiating healthy individuals from schizophrenia patients using these anticorrelated networks. The sample consisted of 56 patients with DSM-IV schizophrenia and 56 healthy controls. We computed functional connectivity matrices and used Anticorrelation after Mean of Antilog method (AMA) to select predominantly anticorrelated networks. The basal ganglia, thalamus, lingual gyrus, and cerebellar vermis showed significantly different, Type A (decreased anticorrelation) connections. The medial temporal lobe and posterior cingulate gyri showed significantly different, Type B (increased anticorrelation) connections. Use of SVM on AMA networks showed moderate accuracy in differentiating schizophrenia and healthy controls. Our results suggest that anticorrelated networks between the sub-cortical and cortical areas are abnormal in schizophrenia and this has potential to be a differential biomarker. These preliminary findings, if replicated in future studies with larger number of patients, and advanced machine learning techniques could have potential clinical applications.

Keywords: Anticorrelated network; Machine learning; Resting state fMRI; Schizophrenia; Support vector machine; Thalamus.