Abstracts

Rationale: For patients with medically refractory temporal lobe epilepsy (TLE), the most effective treatment remains surgical resection of the epileptogenic zone (EZ) via anterior temporal lobectomy (ATL). Successful outcomes from resective surgery, particularly in non-lesional cases, depend on accurate identification of the EZ and inclusion of the EZ within the surgical resection. We aimed to compare the results of a network fragility algorithm, based on dynamical systems stability theory, applied to electrocorticography (ECoG) recordings to those of standard clinical ECoG interpretation with respect to EZ localization and post-surgical seizure-freedom rates. Methods: All patients who underwent phase-two monitoring with subdural grids and/or depth electrodes prior to ATL for medically refractory TLE at the University of Miami from 2/1/20014 – 2/1/2018 were identified and a subset of these with available adequate ECoG and imaging data were selected for inclusion in the study (n=4). Demographic and clinical patient data were collected, including seizure freedom rates. Subsequent analyses of ECoG and imaging studies were conducted to generate a fragility heat map of the proposed EZ for each patient. Each heat map displays how likely (i.e. fragile) each ECoG contact is at each time point leading up to the time of seizure onset. Results: A fragility heat map for each patient was generated that shows the fragility of ECoG contacts over time. The more red a spatio-temporal region was, the more likely the algorithm determined that contact to belong to the EZ, while the blue regions indicated a likely normal tissue region. Of the four patients, two had Engel class I outcome whereas two had seizure recurrence (Engel class II and III respectively). In both successful outcomes, the algorithm highlighted fragile contacts were within the surgically resected region. In both failed outcomes, the algorithm highlighted fragile contacts were outside the region that was included in the ATL, which suggested that the resection did not include all regions implicated to be part of the EZ. Specifically, patient 1 (Engel score I) only had fragile electrodes within the right anterior temporal lobe which was resected. In patient 3 (Engel score III), the standard ATL did not include the basal occipital region which the algorithm predicted to be part of the EZ. Conclusions: The novel network fragility algorithm accurately predicted the EZ with patients that had removal of all fragile electrodes remaining seizure free whereas patients where fragile electrodes were predicted outside the area surgically resected experiencing seizure recurrence. More specifically, the heat maps had a high/low degree of agreement with clinical annotations of the EZ in patients with successful/failed surgical outcomes. This tool may enable clinicians to identify the seizure focus in patients with medication refractory TLE and guide surgical interventions to yield improved clinical outcomes. Funding: No funding