Fig. 6: Similarity between structural (log normalized streamline count) and functional connectivity estimates.

The detection accuracy (measure via AUC) of structural connectivity between brain regions using several functional connectivity estimation methods, or a representative patient (#5) using three recording montages -- referential (a), local multiple linear regression (c), and global mean signal regression (e) montages. b, d, f The structure-function coupling, measured as the correlation between structural and functional connectivity weights of sampled brain regions for the same patient and three recording montages. Functional connectivity estimates based on the pairwise MEM and co-activation rates were calculated from binarized (‘0’ and ‘1’) power amplitude states. The remaining functional connectivity estimates (Pearson correlation, partial correlation, PLV, and WPLI) were derived from the band-passed iEEG and band-passed iEEG power time series. The results for different frequency bands are color-coded based on the legend in panel e. The AUC and correlation values significantly higher than those of geometric structural nulls (one-sided permutation test, n = 10, 000, p < 0.05) are depicted as ‘X’, and values significant after FDR correction for multiple comparisons across frequency bands are depicted as ‘O’.