Abstracts

Rationale: Electrophysiological studies have revealed pathological processes involved in epileptiform activity, but the role of macro-scale networks in seizure initiation, propagation and termination remains unclear. Intracranial electroencephalography (iEEG) studies have shown desynchronization of seizure foci from surrounding tissue prior to the onset of a seizure, but the changes in network structure/function during seizure propagation and termination have not been characterized. In this study, we use model-based network estimation to identify and quantify network state changes during seizures to better understand this process.

Methods: iEEG data from 10 epilepsy patients from the iEEG.org database were analyzed. Criteria for inclusion were availability of complete clinical notes, seizure freedom post-surgery (ILAE Class I) and the presence of clearly defined and spatially limited seizures. Recordings for each patient were common mode average referenced and bandpass filtered (0.5-500Hz). Each clinical seizure was divided into three 15-second periods, initiation, mid-seizure, and termination, and each period was further divided into consecutive 5-second epochs. In each epoch, a multiple input, single output (MISO) state space model was estimated for each channel output with all other channels as inputs. The amount of influence of each other channel on the given channel was used to infer a directed network graph of the relationship between all channels for each time window. iEEG contacts were represented as nodes and assigned to 3 clinically-determined regions: seizure onset zone (SOZ), within 2 contacts of SOZ (PeriSOZ or PSZ) and other non-SOZ (NSZ). The resulting networks were analyzed across seizures and patients using degree centrality, an index of the proportion of directed connections through each node. We also analyzed equivalent periods of interictal data (at least 2 hours before or after seizure) as controls.

Results: Degree centrality in all 3 ictal periods was significantly higher than interictal in all regions (Fig. 1). Surprisingly, by the mid-seizure period, SOZ degree fell significantly below both PSZ and NSZ groups, but rose again during termination, with distant channel degree falling significantly in this time period. This counterintuitive result may be explained by examining the proportion of incoming vs outgoing connections (Fig. 2). During Initiation, the vast majority of connections in the SOZ are outgoing, projecting to PSZ channels. This falls to a roughly 50/50 incoming vs outgoing during the mid-seizure period. Finally, during termination, SOZ connections are mostly incoming from the PSZ group. Together, this suggests a combined SOZ exhaustion and neighboring inhibition mechanism for seizure termination.

Conclusions: MISO model network estimation identified temporal changes in iEEG network properties. Analysis of network structure provides quantitative computational evidence from human data for the theory that seizure termination results from a combination of SOZ exhaustion and surround inhibition.

Funding: Please list any funding that was received in support of this abstract.: Supported by UConn Health Dept of Neurology.