Impact of Negative Correlations in Characterizing Cognitive Load States Using EEG Based Functional Brain Networks

The human brain is one of the least understood large-scale complex systems in the universe that consists of billions of interlinked neurons forming massive complex connectome. Graph theoretical methods have been extensively used in the past decades to characterize the behavior of the brain during different activities quantitatively. Graph, a data structure, models the neurophysiological data as networks by considering the brain regions as nodes and the functional dependencies computed between them using linear/nonlinear measures as edge weights. These functional connectivity networks constructed by applying linear measures such as Pearson’s correlation coefficient include both positive and negative correlation values between the brain regions. The edges with negative correlation values are generally not considered for analysis by many researchers owing to the difficulty in understanding their intricacies such as the origin and interpretation concerning brain functioning. The current study uses graph theoretical approaches to explore the impact of negative correlations in the functional brain networks constructed using EEG data collected during different cognitive load conditions. Various graph theoretical and inferential statistical analyses conducted using both negative and positive correlation networks revealed that in a functional brain network, the number of edges with negative correlations tends to decrease as the cognitive load increases.