Abstracts

Stimulation-induced Seizures Map Spontaneous Seizure Spread Networks

Abstract number : 2.159
Submission category : 3. Neurophysiology / 3E. Brain Stimulation
Year : 2024
Submission ID : 895
Source : www.aesnet.org
Presentation date : 12/8/2024 12:00:00 AM
Published date :

Authors :
Presenting Author: William Ojemann, BS – University of Pennsylvania

Akash Pattnaik, BS – University of Pennsylvania
Andrew Revell, MD, PhD – University of Pennsylvania
Eli Cornblath, MD, PhD – Unviersity of Pennsylvania
Joshua LaRocque, MD, PhD – University of Pennsylvania
Jacob Korzun, MD – University of Pennyslvania
Catherine Kulick, MD – University of Pennsylvania
Daniel Zhou, MD – University of Pennsylvania
Kathryn Davis, MD – University of Pennsylvania
Brian Litt, MD – University of Pennsylvania
Erin Conrad, MD – University of Pennsylvania

Rationale:
Evaluation for epilepsy surgery is often lengthy, requiring clinicians to wait up to weeks for spontaneous (spont) seizures during intracranial monitoring. During this period, clinicians can induce seizures using low-frequency stimulation, which have been shown to start in tissue that, if resected, portends good clinical outcomes. However, the extent to which stimulation-induced (stim) seizures map the same epileptic networks as spont seizures remains unknown. Here, we compare these networks in a group of patients during evaluation for epilepsy surgery at the University of Pennsylvania (F1A).




Methods: We examined stim and spont seizures from 16 patients during stereo-EEG recording. To accurately measure seizure onset and spread we compared four unsupervised or pre-trained models – two deep learning and two feature-based – to consensus manual annotations of seizure onset and 10 second spread on a subset of seizures per-patient (F1B,C). Using predictions from the highest performing model, we defined seizure similarity for all pairs of seizures within patients to be the Cohen’s Kappa (F1D) at seizure onset at the channel and region level localized to the DKT atlas (F1E). To determine if stim seizures map spont seizure spread regions, we measured the percentage of stim seizure onset regions to which the spont seizure spreads over the course of the seizure. We compared this percentage to a negative control of random regions selected from the stim seizure as well as a positive control of the percentage of spontaneous onset regions that overlap with each other averaged across patients.


Results: Absolute slope was the highest performing model for identifying seizure onset and 10 second spread channels at a threshold of 0.75 (F1C). Seizure onset similarity was significantly higher between spont seizures than between stim and spont seizures at both the channel (t-test, p < 0.0001
Neurophysiology