The progress in predictive modeling of post-stroke epilepsy

Post-stroke epilepsy (PSE) is a significant complication of both ischemic (IS) and hemorrhagic strokes (HS), leading to increased morbidity and reduced quality of life. Accurate prediction of PSE risk is essential for early intervention and tailored management. Multiple predictive models have been developed for different stroke subtypes. In HS, models such as CAVE, CAVS, CAV+, and CAVE2 emphasize lesion characteristics and early seizures. Within IS, models such as SeLECT and PSEiCARe focus on cortical involvement, large-artery atherosclerosis, and early seizure occurrence. Recent advances in machine learning-based approaches have shown improved predictive accuracy for both IS and HS patients, although further validation is required for routine clinical application. This review summarizes and compares predictive models for PSE across stroke subtypes, highlighting their clinical relevance and potential for improving patient outcomes through early risk stratification. Integration of multimodal data may further enhance seizure prediction and guide personalized intervention strategies.