PhD Student University of Pennsylvania Philadelphia, Pennsylvania
Rationale: Despite the success of responsive neurostimulation (RNS) therapy for epilepsy, demonstrated in a recent post-approval real-world study by a median seizure frequency reduction of 67% and 82% at one and three years respectively , it is evident that not all patients are good candidates for RNS. Additionally, while a patient is considered to be an RNS responder if they show at least a 50% reduction in seizure frequency compared with baseline, there is still a need to improve spatial stimulation targeting to enhance seizure suppression in these patients. Neuroimaging and intracranial EEG (iEEG) recordings collected from patients in the epilepsy monitoring unit (EMU) prior to RNS surgery provide valuable data for identifying predictive biomarkers that can (1) identify patients who are most likely to respond to RNS, and (2) designate favorable locations for RNS lead implantation. Methods: We validated a data analysis pipeline for discovering biomarkers that demonstrate the potential to guide RNS therapy decisions. We performed preliminary evaluation of two candidate biomarkers and replicated our framework across three major epilepsy centers using EMU data from 30 RNS patients who had pre-implant iEEG and at least one year of post-RNS follow up. First, we generated functional networks across sequential time windows of iEEG seizure recordings, and quantified how peri-ictal changes in synchronizability, a network metric that has been shown to accurately predict outcome from resective epilepsy surgery, could distinguish between good and poor RNS responders. Next, we quantified the distance between RNS electrode contacts and iEEG contacts at the seizure onset zone (SOZ) using a patient-specific co-registration procedure, and examined how this distance correlated with a reduction in seizure frequency. Results: We found that patients who maintained an elevated network synchronizability after seizure onset were classified as RNS responders at the most recent visit, a result that runs contrary to synchronizability predictions in resection patients. We also found that the distance of RNS leads from SOZ had no significant correlation with the patient’s reduction in seizures. Conclusions: Our findings support the quantitative, spatiotemporal analysis of iEEG in the context of planning RNS therapy and highlight that stimulating seizure-generating regions with the intention of interrupting emerging seizures may not be the primary driver of patient response. We share all annotated intracranial EEG and imaging data to facilitate ongoing biomarker discovery for guiding RNS therapy.
 Lin, et. al. Real-world experience with brain-responsive neurostimulation for focal onset seizures. Poster, AES 2019 (Abst. 1.216) Funding: Please list any funding that was received in support of this abstract.: We would like to acknowledge support from the Mirowski Family Foundation, Johnathan & Bonnie Rothberg, and Neil & Barbara Smit.