Capturing Cognition via EEG-Based Functional Brain Networks

Abstract

The human brain is comprised of complex networks of neuronal connections, with the functioning of these networks underscoring human cognition. At any given point in time, the complexity of these networks may be greater than the entire communications network on the planet yet functional brain networks are not static; instead, they form and dissolve within milliseconds. Although much is known about the functions and actions of individual neurons in isolation, at a systems level, when billions of neurons coordinate their individual activity to create functional brain networks and thus cognition, understanding is limited. This is due in part to the system behaving completely differently to its parts; that is, emergent properties such as intelligence, emotion and cognition cannot be adequately explained from a sum-of-parts perspective; what is needed instead are powerful computational techniques to model and explore both the intricacies and dynamics of functional brain networks. Although unravelling the activity of the human brain remains circumscribed by technological and ethical constraints, complex network analysis of EEG data offers new ways to quantitatively characterize neuronal cluster patterns. This, in turn, allows the analysis of functional brain networks to understand the complex architecture of such networks. Despite the increasing attention that functional brain network analysis is gaining in computational neuroscience, the true potential of such analysis to reveal dynamic interdependencies between brain regions has yet to be realized. To address this, multi-channel EEG data has been used to examine the dynamics of such networks during cognitive activity using Information Theory based nonlinear statistical measures such as transfer entropy. Results across different paradigms requiring different types of cognitive effort clearly suggest that transfer entropy is a highly sensitive measure for detecting cognitive activity. Furthermore, these results demonstrate that transfer entropy has clear potential for developing cognitive metrics based on complex features such as connectivity density, clustering coefficient and weighted degree. These techniques may also have application in the clinical diagnosis of cognitive impairment as well as providing new insights into normal cognitive development and function.