Identification of Schizophrenia using Functional Connectivity and Graph Theory through Resting State Functional Magnetic Resonance Imaging (rs- fMRI)

Existing traditionally diagnostic methods for schizophrenia lack objectivity, necessitating the exploration of alternative approaches. Functional connectivity and graph theory analysis using Functional Magnetic Resonance Imaging (fMRI) present promising avenues for developing reliable diagnostic tools. This study investigates the application of functional connectivity and graph theory in fMRI to establish objective and accurate means of early schizophrenia detection. In the present study, Resting State Functional Magnetic Resonance Imaging (rs-fMRI) data was acquired from the Consortium of Neuropsychiatric Phenomics at UCLA. Preprocessing involved utilizing the Automated Anatomical Labeling (AAL) atlas to segment the brain into 90 regions related to cerebral cortex, limbic system, and subcortical structures. Mutual information was employed to compute the functional connectivity matrix, which served as the foundation for constructing a brain graph network, emphasizing significant and robust connections. Five global network features, namely average strength, eccentricity, local efficiency, clustering coefficient, and transitivity, were extracted. A support vector machine (SVM) classifier was then employed to differentiate healthy individuals $(n=50)$ from schizophrenia patients $(n=50)$ based on these influential features. Results demonstrated that the mutual information method combined with the extracted global network features achieved noteworthy performance metrics: $83 \%$ accuracy, $85 \%$ sensitivity, and $92 \%$ specificity using the SVM classifier. The fusion of brain graph features and functional connectivity derived from rs-fMRI data analysis exhibits the potential to accurately identify individuals with schizophrenia.