Dynamic Network Drivers of Seizure Generation, Propagation and Termination in Human Neocortical Epilepsy
Fig 2
Schematic for identifying network configuration states.
(a) We estimate dynamic functional connectivity; colors represent arbitrary connection strengths ranging from strong to weak (red, gray, blue). (b) We track all unique functional connections over time using a configuration matrix, in which each vector represents the set of connection weights for a 1s time window. (c) We compute the similarity between the network geometries of each pair of time windows using a Pearson correlation coefficient. In the resultant configuration-similarity matrix, colors represent the magnitude of similarity and visually identified clusters are distinguished by colored, dashed lines (orange and green). (d) We optimize a modularity quality function to cluster the configuration vectors (and thus time windows) into communities. Each cluster or community contains time windows with similar network geometry; colors represent assignments of time windows to different network configuration communities (orange and green).