Abstracts

Identifying Epileptogenic Regions Through Delayed Responses Evoked from Single Pulse Electrical Stimulation

Abstract number : 3.166
Submission category : 3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year : 2023
Submission ID : 890
Source : www.aesnet.org
Presentation date : 12/4/2023 12:00:00 AM
Published date :

Authors :
Presenting Author: Sayantika Roy, BS – Johns Hopkins University

Golnoosh Kamali, PhD – Johns Hopkins University; Mark Hays, PhD – Johns Hopkins University; Nathan Crone, MD – Johns Hopkins School of Medicine; Sridevi Sarma, PhD – Johns Hopkins University; Rachel Smith, PhD – University of Alabama at Birmingham; Joon Kang, MD – Johns Hopkins School of Medicine

Rationale: Epilepsy is a neurological disease that affects millions of people worldwide. thirty percent have drug-resistant epilepsy, where surgical resection is a common therapy. Successful surgical outcomes, however, are dependent on accurate identification of the epileptogenic zone (EZ), but current methods to identify the EZ have poor success rates. A new method of localizing the EZ is through analysis of the evoked responses from single pulse electrical stimulation (SPES). Responsive channels produce consistent responses after stimulation, called early responses. However, in certain channels, random spikes in electrical activity can occur, called delayed responses (DR). These may correlate to the epileptogenic regions of the brain. Despite this promising clinical utility, analysis of DRs has not been widely adopted due to the amount of time and difficulty required to identify these responses with variable timing and morphology. We hypothesize that these responses can be automatically detected and measured to be used as markers of EZ that translate to important surgical targets.

Methods: We retrospectively analyzed intracranial EEG (iEEG) data from five patients with medically refractory epilepsy (MRE) that underwent SPES at the Johns Hopkins Hospital. We visually inspected the iEEG data for the presence of DRs, which was verified by a board-certified epileptologist. Using our visually identified DR dataset, we developed a delayed response detection algorithm that identified possible occurrences of DRs within a window of 300-1200ms after stimulation onset. Detection criteria was an amplitude threshold based on the mean and standard deviation of each recording channel, optimized for high sensitivity and specificity.

Results: We found that using an average threshold of 1.06 standard deviations from the mean proved optimal on identifying occurrences of DRs. With this threshold, we were able to detect the presence of DRs with specificity and sensitivity of 70% and 68%, respectively, and 70% accuracy. 

Conclusions: We visually analyzed and identified DRs in 5 patients with focal MRE and developed a patient-specific amplitude threshold detector that can automatically identify for the presence of DRs. Through the creation of a quantitative and objective methodology, our introduction of an automated DR detector can aid in the adoption of an automated biomarker. With the inclusion of more datasets and refined parameters to improve detection rates, along with validating the correlation with successful surgical outcomes, we will be able to provide an automated biomarker to aid clinicians in localizing the EZ, resulting in improved patient surgical success rates which will lead to improved clinical outcomes and quality of life.

Funding: No funding was used to support this project. 

Neurophysiology