High-Efficiency Graph Measures for Discriminating Schizophrenia Patients from Healthy Controls Using Structural and Functional Connectivity

Mahya  Naghipoor-Alamdari; Jafar  Zamani; Farzaneh  Keyvanfard; Abbas  Nasiraei-Moghadam

doi:10.18502/fbt.v12i2.18282

Mahya Naghipoor-Alamdari Department of Biomedical Engineering, Amirkabir University of Technology, Tehran, Iran
Jafar Zamani School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
Farzaneh Keyvanfard School of Cognitive Sciences, Institute for Research in Fundamental Sciences, Tehran, IranSchool of Cognitive Sciences, Institute for Research in Fundamental Sciences, Tehran, Iran
Abbas Nasiraei-Moghadam Department of Biomedical Engineering, Amirkabir University of Technology, Tehran, IranDepartment of Biomedical Engineering, Amirkabir University of Technology, Tehran, Iran

DOI: https://doi.org/10.18502/fbt.v12i2.18282

Keywords: Schizophrenia; Independent Component Analysis; Subnetworks; Functional Connectivity; Structural Connectivity; Graph Theory.

Abstract

Purpose: Schizophrenia (SZ), which affects 0.45% of adults worldwide, is a complex mental illness with unknown causes and mechanisms. Neuroimaging techniques have been used to study changes in the brain of patients with SZ. In this study, we aim to construct brain subnetworks, analyze the association of structure with function, and investigate them with graph measures. We hope to identify important subnetworks and graph measures for SZ diagnosis.

Materials and Methods: This study investigates the structural and functional brain connectivity of 27 Healthy Controls (HC) and 27 patients with SZ. Independent Component Analysis (ICA) and joint ICA (jICA) are used to construct subnetworks based on functional and structural connectivity. An association between structural and functional connectivity is examined. Joint functional and structural subnetworks are also examined and compared with independent analysis of functional and structural subnetworks. Several graph measures are used in the whole brain and its subnetworks.

Results: In this study, we investigated brain connectivity in HC and SZ patients using graph measures. The study analyzed both the whole brain and brain subnetworks to better understand the importance of partitioning the brain into subregions. Our results suggest that analyzing the whole brain may not be the most effective method for studying the brain peculiarities of SZ patients. In addition, multimodal brain analysis has proven to be effective in understanding SZ. There is no one-to-one relationship between structural and functional connectivity in the brain. Certain measures such as maximum modularity, clustering coefficient, network strength, global efficiency, and path length were important in distinguishing patients with SZ from HCs in specific subnetworks. This study recommends further investigation of specific subnetworks that overlap with default mode, visual, and somatomotor resting state networks.

Conclusion: This study emphasizes the importance of subnetwork and multimodal analysis for understanding SZ disease.