Fig. 4: Schematic of the dynamic functional connectivity processing.

A Sliding-window functional connectivity maps were calculated in 30-s time intervals (1-s step size), and unsupervised clustering was performed via k-means to identify prominent clusters (brain states). B Brain states identified by k-means clustering and sorted based on the mean correlation within each state. C Results of the leave-one-group-out validation showing that k-means yields equivalent clusters in all the five conditions. The connectivity maps for all the conditions are in Supplementary Fig. 6.