Fig. 7: Schematic diagram of dynamic functional connectivity and brain state analysis.

BOLD time series were first extracted for each region of interest. Phase-based dynamic functional connectivity (DFC) was computed by generating instantaneous connectivity matrices at each time point. The leading eigenvector of each matrix was then derived, representing the dominant connectivity pattern at that moment. These eigenvectors were concatenated across all subjects to form a unified dataset. K-means clustering using Manhattan (L1) distance and 50 iterations was applied to identify recurring connectivity configurations, termed brain states. Finally, the temporal characteristics of these brain states, including their probability of occurrence, were quantified.