Fig. 2.
Distributed connector hubs are re-organized with arousal modulations during resting state. (a) The total number of resting state networks (N) detected by SPARK from individuals were preserved between high and low arousal states. (b and c) The group average k-hubness maps at high (b) and low (c) arousal. (d) The map of difference in the group average k-hubness between the low and high arousal states. (e) The estimation of group-level HDI<k> between high and low arousal. A linear regression model is used to find a linear fit of nodal group-average k-hubness (<k>) estimated from the two states. HDI<k> is defined as a slope of the linear fit. (f) The estimation of group-level HDI<k> between two randomized states, by averaging k-hubness across 702 false brain-pupil pairs in each node. (g) An example of individual-level HDIk from a single subject exhibiting the median of HDIk within group. Note that nodal k-hubness is an integer, therefore nodes with a same value are superimposed in this scatter plot. (h) The distribution of individual-level HDIk (top) and those from null data (bottom). p-value estimated using the left-tailed Wilcoxon rank sum test is shown. (i) The bar plot of k-hubness distributions within the eleven pre-defined large-scale networks in each state. Mean ± standard deviation. (j) Data points in figures e, f and j are color-coded using eleven a priori functional networks. Ten networks were defined as described in Noble et al. (2017), and the nodes belonging to the brainstem were assigned to an 11th network.