Investigation of linear and non-linear functional connectivity within resting-state networks using graph theory in Parkinson's disease

Abstract

Purpose Parkinson's disease (PD) is widely known as a neurodegenerative disorder of the nervous system for which there is no cure. Accordingly, researchers can utilize neuroimaging techniques like functional magnetic resonance imaging (fMRI) to investigate neural activities in the brain non-invasively. Most previous research works construct brain graphs based on linear correlations for functional connectivity (FC) analysis. In this study, we compared linear and nonlinear functional connectivity methods. Methods The objective of our study is to implement 5 functional connectivity methods on 14 resting-state fMRI networks (RSNs) based on the FIND RSN template that is divided into 90 regions. Kernel Mutual information (KMI), a unique nonlinear connectivity approach based on Mutual information (MI), is also employed. Consequently, the validity of the methods was assessed using local graph measures and statistical analysis. Results The results show that nonlinear methods outperformed linear ones using the outcome of graph theory. In the non-linear functional connectivity methods, all seven graph measures showed a significant difference between two groups: healthy control (HC) and Parkinson's disease (PD), but only one graph measure showed a significant difference in the linear functional connectivity methods. Furthermore, while K-Corenness centrality has been utilized in previous studies to diagnose and assess various neurodegenerative illnesses, it is employed for the first time in our study to diagnose Parkinson's patients using fMRI data. Conclusions According to the findings of this study, nonlinear functional connectivity should be investigated in Parkinson's disease and other neurodegenerative diseases.