Fig. 2: Resting-state and task-fMRI sessions contain reoccurring FNC states.

a Four representative FNC states (cluster centroids) were derived from the k-means clustering applied to sliding-window FNC matrices from the resting-state (upper row) and three task-fMRI sessions (lower row). Each FNC state represents a reoccurring functional connectivity pattern between 61 ICA components, which are categorised into seven functional domains (Fig. S2). The percentage of each state within a scanning session is shown in parentheses. (Note: Due to decimal approximation, the percentages for the four FNC states in the EFT and MID sessions sum to 100.01% and 99.99%, respectively). Connectivity profiles of the same state across resting-state and task-fMRI sessions exhibit a higher correlation (average r = 0.86 ± 0.17) compared to correlations between different states (average r = 0.67 ± 0.10, two-sample t-test: t₁₁₈ = 7.17, p = 6.93 × 10⁻¹¹, 95% CI = [0.13, 0.24], Coden’d = 1.63). b The four FNC states differ in network modularity and participation coefficients, which quantify the level of segregation and integration between functional modules. The modularity and participation coefficients are averaged across resting-state and task-fMRI sessions. c One-way ANOVA and post-hoc t-test comparisons between FNC states on modularity and participation coefficients. Red lines indicate significant differences between FNC states after FDR correction (Table S4). Abbreviations for the seven functional domains: subcortical (SCN), temporal (TEP), sensorimotor (SMN), visual (VSN), cognitive control (CON), default mode (DMN) and cerebellar (CEB) networks.