Abstracts

Rationale: Focal onset neocortical seizures are often characterized by the rapid spread of ictal activity from the seizure onset zone. As such, these patients are typically poor surgical candidates due to the difficulty in discerning the location and extent of the ictal-generating cortex. Recent advances in neuroimaging and multivariate analysis techniques have spurred interest in understanding the complex network interactions which arise during epileptiform events as well as possible targets of therapeutic intervention. In this study, we utilize graph theory analysis techniques to study critical nodes in the epileptogenic networks during periods of ictal activity. Methods: Electrocorticogram recordings were obtained from 25 patients with neocortical onset seizures and were studied under a protocol approved by the Institutional Review Boards at the University of Minnesota and the Mayo Clinic (Rochester, MN). The directed transfer function was applied to each seizure and two graph measures, the outdegree (OD), a measure of the amount of activity originating from each node, and betweenness centrality (BC), a measure of a node’s importance to the flow of information within a network, were calculated from the networks in the theta, alpha, beta and gamma frequency bands. K-means clustering was used to identify activated nodes during the ictal activity and the OD and BC values for these nodes were calculated at 5, 3 and 1 minute prior to and following ictal onset and offset in addition to time points during the early-, mid- and late-ictal periods. A total of 65 seizures were analyzed in this manner. Results: The OD of the activated nodes was significantly increased during the early, mid and late ictal periods as compared to preictal baseline levels while the BC values in the theta, alpha and beta frequencies decreased at the onset of the ictal activity and reached a minimum at ictal cessation. In the gamma band, the BC values were initially increased before later declining, suggesting network connectivity via a gamma-modulated network at ictal onset. Analysis of the spatial extents of the active OD and BC networks revealed a high degree of coincidence with the clinically-identified epileptogenic zones which were frequency-dependent with the gamma and beta networks having a significantly greater amount of overlap as compared to the theta networks. Conclusions: In the study we have examined changes in two graph theory measures during preictal, ictal and postictal periods. For the OD measured in the active nodes, an increase was observed beginning at ictal onset and persisting until ictal cessation. Likewise, a decrease in the BC measured in the theta, alpha and beta frequencies was identified during the course of the seizure while the BC of the active nodes in the gamma frequency band was increased during the ictal initiation. The use of these graph theory measures can provide a useful tool in the identification of the cortical regions responsible for the initiation and propagation of the ictal activity which in turn could lead to more accurate and focal surgical resections in patients with medically intractable epilepsy.