Contribution of structural connectivity to MEG functional connectivity
AbstractStructural connectivity by axonal fiber bundles provides the substrate for transmission of action potentials across the brain. Functional connectivity in MEG signals is expected to arise from communication along structural connections. However, very little empirical evidence has been obtained to support this hypothesis. The main objective of this study is to use simulations and MEG data to directly evaluate the contribution of structural connectivity to MEG functional connectivity measures. Since axonal transmission is on a millisecond time scale we hypothesize that measures sensitive to phase synchronization in a frequency band, such as coherence, would have a closer relationship to structural connectivity than measures sensitive to slower time scales such as amplitude-envelope correlation. We estimate graphical models of MEG functional connectivity, i.e, the MEG effective connectivity, to reduce the influence of leakage effects and common input effects, and to explicitly model the contribution of structural connectivity to functional connectivity. Consistent with our hypothesis, networks defined by models of gamma band (> 30 Hz) coherence that incorporate phase information show the closest alignment to structural connectivity. However, at lower frequencies (1-30 Hz) there was better alignment between models of amplitude envelope correlation and structural connectivity. In simulations, summarizing network properties of graphical models using graph theoretic metrics provides a robust measure of the relationship between functional and structural connectivity that is preserved even at low signal to noise ratios. In MEG data, centrality of nodes in the gamma band networks more closely correspond to centrality of nodes in the structural networks than a direct comparison between edge weights.