A Unified Approach to Post-Stroke Epilepsy (PSE) Prediction Across Stroke Subtypes
Abstract number :
2.279
Submission category :
4. Clinical Epilepsy / 4D. Prognosis
Year :
2025
Submission ID :
1030
Source :
www.aesnet.org
Presentation date :
12/7/2025 12:00:00 AM
Published date :
Authors :
Presenting Author: Justin Wheelock, BA – Yale School of Medicine
Yilun Chen, BS – Yale University
B. Ayvaz, MD – Yale School of Medicine
Sithmi Jayasundara, BS – Yale University
Rachel Choi, BS – Yale University
Tejaswi Sudhakar, BS – Yale University
Adeel Zubair, MD – Yale University
Adithya Sivaraju, MD – Yale New Haven Hospital
Sahar Zafar, MD, MBBS – Massachusetts General Hospital
Aaron Struck, MD – University of Wisconsin, Department of Neurology, Madison, WI
Lawrence Hirsch, MD – Yale University School of Medicine
Emily Gilmore, MD – Yale School of Medicine
M. Brandon Westover, MD, PhD – Beth Israel Deaconess Medical Center
Jennifer Kim, MD, PhD – Yale School of Medicine
Rationale: Epilepsy is a morbid complication affecting patients of all stroke types: acute ischemic stroke (AIS), intracerebral hemorrhage (ICH) and subarachnoid hemorrhage (SAH). While PSE risk factors have been described, the expertise required for specialized scoring systems limit broad implementation. We aimed to identify common PSE risk factors across stroke subtypes that could be automatically extracted from electronic health record (EHR) data. We evaluated PSE prediction using (1) routine clinical assessments (2) qualitative imaging and epileptiform patterns on continuous EEG (cEEG), and (3) automated, quantitative lesion volumes from computed tomography (CT) and magnetic resonance imaging (MRI).
Methods: We identified a retrospective cohort of adult (age ≥ 18) AIS, ICH, and SAH patients admitted to a tertiary care center (2014-2022) with available imaging and continuous EEG ≤7 days from stroke onset. PSE was defined as seizure >7 days from stroke onset. We excluded patients with a history of seizures and other epileptogenic condition (prior brain injury, lesion, metastases). We extracted clinical features, including assessments, surgical treatments, and inpatient complications, from the EHR as well as qualitative imaging and EEG features from radiology and neurophysiology reports, respectively. Quantitative imaging features were extracted using automated tools for CT hemorrhage volume and MRI infarct volume. Our primary outcome was PSE occurrence, and patients were censored at death or loss to follow-up. We used Cox Proportional Hazard (CPH) modeling to identify PSE predictors, and selected features with p< 0.1 to include in a series of
Clinical Epilepsy