Test of robustness of fractal dimension in differentiating higher-cognitive (HCFNs) and perceptual (PNs) brain networks

Intrinsic brain activity is organised into large-scale networks displaying specific structural-functional architecture, known as resting-state networks (RSNs). RSNs reflect complex neurophysiological processes and interactions, and have a central role in distinct sensory and cognitive functions. The characterization of RSNs in terms of their anatomical and functional properties is therefore crucial to understand the mechanisms of brain organization and the behaviour it supports. In this study, we tested the hypothesis that different RSNs might display different brain dynamics. To this end, we investigated how the fractal dimension (FD) of RSNs derived from high density electroencephalography (hdEEG) at rest can be used to differentiate RSNs, and how the FD property of RSNs is linked with their functional roles. Moreover, we tested the robustness of the FD measure respect to different parameters used for estimating the FD such as segment length, overlapping and k-value. We were able to identify two clusters of RSNs, one named perceptual network mainly consisting of sensory networks (PN, including auditory, sensorimotor and visual networks) and the other more related to higher cognitive functions (HCFN, including default mode network and attention network) independently from the parameter chosen to estimate the FD. These clusters were defined in a completely data-driven manner using hierarchical clustering. We concluded that, FD is a valid method for assessing resting brain functions by providing valid information on the mechanisms underlying normal brain activation and that this method is a robust and easy approach to recognize mechanisms underlying brain activity that could, so far, be overlooked by the more classic linear methods