Fig. 8.
Global signal regression has an impact on hub detection using Graph theory across arousal levels. For each individual subject, participant coefficient (PC) was calculated for each node in the shen-268 functional atlas from the weighted undirected network (PCW) and the binary undirected networks constructed using the proportional threshold 30% (PC30%). Group average participant coefficient (⟨PC⟩) was computed by averaging PC across subjects in each node. We compared the distributions of between-state changes in group-average PC (Δ⟨PC⟩, low-high) within each of the 11Net pre-defined large-scale networks (color-coded). The null distribution of Δ⟨PC⟩ was generated from the same nodes in each network over 5,000 permutations. Color-coded asterisks indicate Bonferroni corrected p-values from the two-tailed Wilcoxon rank sum tests, *: p<.05, **: p<.01, ***: p<001. This figure shows the summary of network-level Δ<PC> distributions using the mean of Δ⟨PC⟩ within each network from arousal state-stratified datasets preprocessed (a and b) with GSR and (c and d) without GSR. (e and f) shows the comparisons between the results with and without GSR. p-value is presented on top of each comparison using the two-tailed Wilcoxon rank sum test.