Abstracts

Rationale: Functional architecture of the epileptic neocortex has been studied extensively to better identify the optimal target for surgical resection and, more recently, implantable devices. However, the optimal network target for a chronic device is elusive and requires further dissection of how seizures evolve within the epileptic network. Biomarkers of epileptiform activity, such as epileptic spikes and high-frequency oscillations, believed to circumscribe optimal resection regions, often only partially overlap with the seizure-generating network and poorly relate to patient outcome. Seizure dynamics are often characterized by stages with varying degrees of synchronization. Could the efficacy of resective surgery be improved by controlling network synchronization? Specifically, we ask, “How would network synchronizability respond to virtual resection of specific brain regions (or network nodes)?” Furthermore, can we use this method to pinpoint control regions that regulate network synchronizability?Methods: We constructed time-evolving functional networks from electrocorticography recorded from 6 patients diagnosed with drug-resistant neocortical epilepsy undergoing routine pre-surgical evaluation through the International Epilepsy Electrophysiology Portal (IEEG Portal, http://www.ieeg.org). We analyzed seizure and pre-seizure epochs. In each epoch we divided EcoG signal into 1s non-overlapping time-windows and estimated functional connectivity in gamma (30–40 Hz) and high-gamma (95–105 Hz) frequency bands using multitaper coherence estimation. We assessed the impact of virtually resecting each node from the network over all time-windows by computing a novel control centrality metric. The control centrality of a node is the fraction change in synchronizability as a result of removing the node from the network. Positive (negative) values of control centrality indicate the node has a desynchronizing (synchronizing) influence on pre-resection synchronizability.Results: We computed the control centrality over 16 secondarily generalized complex partial seizures. In gamma and high-gamma functional networks, we observed a few nodes that either substantially increased (desync nodes) or decreased (sync nodes) synchronizability. Moreover, these influential controllers were not necessarily seizure-generating nodes. Furthermore, we found that strong controllers outside the seizure-generating network exhibited unique changes in desynchronizing and synchronizing roles preceding and during seizures.Conclusions: Virtual resection of the epileptic network is a novel and powerful method to probe network response to targeted surgical resection. Using virtual resection, we identified influential nodal controllers of network synchronizability that occupy network regions away from the seizure-generating network. Future work will validate the mechanistic role of controllers in regulating spatial extent of seizure evolution.