Item

Abstract

Introduction: Are functional network topologies of the brain candidates for a neural account of sensory awareness? Pre-stimulus network topologies have been reported to predict if tactile or auditory stimuli enter consciousness or not (Weisz et al., 2014; Sadaghiani et al., 2015). Post-stimulus network topologies have been shown to vary with awareness/unawareness of visual stimuli and to have explanatory power beyond local BOLD amplitudes and baseline functional connectivity (Godwin et al., 2015). Here we test to what extent functional brain networks vary with somatosensory awareness. Methods: We acquired fMRI on a Siemens MAGNETOM Prisma 3 T (TR=750ms; TE=25ms; flip angle (FA)=55°; FOV=192x192mm; 36 3-mm axial slices with 0.5-mm gap; in-plane resolution 3x3mm) while participants (N=38) had to report the perception of single near-threshold electrical pulses (pulse width=0.2ms) applied to their left index finger and their confidence about this yes/no decision. Functional networks were modelled using the generalized psychophysiological interaction (gPPI, McLaren et al., 2012) without the deconvolution step (O'Reilly et al., 2012) for two sets of nodes: (a) a data-driven network of 19 nodes in frontal, parietal and somatosensory cortices and (b) a whole-brain resting-state fMRI atlas of 264 nodes (Power et al., 2011). The 19-nodes network of interest was derived from the BOLD signal contrasts between aware, unaware and control trials without stimulation. gPPI has the advantage of controlling its context-dependent functional connectivity estimates for the BOLD response due to the stimulation and the baseline functional connectivity across the experiment. Graph theoretical analyses of the functional networks were performed with Brain Connectivity Toolbox (Rubinov & Sporns, 2010) across a range of proportional network thresholds. Results: The 19-nodes network showed an increased modularity with somatosensory awareness (60%-thresholded network aware vs. unaware; Wilcoxon's signed-rank test, P = 0.016, probability of superiority PS = 0.71; Fig. 1A). Modularity is a measure of global segregation into distinct networks. Measures of local segregation (clustering), integration (path length) and centrality (participation) showed no consistent significant differences. This indicates that primarily the global pattern of distinct networks within the network of interest changed with stimulus awareness. The same analysis was repeated for the 264-nodes whole-brain network and resulted in no significant differences between aware and unaware trials in modularity, participation, clustering and path length. When the connectivity strengths between the 19 nodes was directly compared with a paired two-sided t-test, aware compared to unaware trials showed a higher intra-parietal and parietal-frontal connectivity, along with a two-fold pattern for somatosensory cortex: (1) primary somatosensory cortex had an increased and (2) secondary somatosensory cortex a decreased functional connectivity to parietal cortex (Fig. 1B). Conclusions: Our results indicate that somatosensory awareness is accompanied by functional brain network topology alterations. These network topology alterations can be observed between functionally relevant nodes that form more modules during somatosensory awareness, indicating functional network segregation and specialization.