Identifying Influential Nodes in a Network Model of Epilepsy

A significant proportion of individuals with epilepsy suffer from intractable forms of the disease. Current evidence suggests that pathological brain connectivity could be a major contributor to the propagation of focal seizures, constituting a possible cause for some forms of intractable epilepsy. Currently, however, the precise network structures that underpin epileptic brain connectivity are poorly understood. In this study, we use a computational model to simulate focal seizure spread in the macaque cortical connectome. We then use the results to propose a novel network centrality measure (called “Ictogenic Centrality”) that accurately identifies which nodes are most effective in propagating seizures. In the framework presented, ictogenic centrality outperforms other standard centrality measures in correctly identifying ictogenic nodes, exhibiting high accuracy (0.947), specificity (0.939), and sensitivity (0.964). Ictogenic centrality is degree based and relies on only a single free parameter, making it useful and efficient to compute for large networks. Our results suggest that baseline brain connectivity may predispose the temporal and frontal lobes toward ictogenicity even in the absence of any overtly pathological network reorganization.