Development and International Validation of a Novel Imaging-Based Risk Score (IsCHEMiA) for the Prediction of Poststroke Epilepsy

BACKGROUND AND OBJECTIVES: Stroke is one of the most common causes of adult-onset epilepsy. We aimed to develop a model to predict poststroke epilepsy (PSE) after a first-ever ischemic stroke, incorporating neuroimaging features of incident stroke. METHODS: We analyzed clinical and neuroimaging features of patients with first-ever acute ischemic stroke consecutively admitted to Massachusetts General Hospital, United States. We performed competing risk regression with all-cause mortality as a competing event and derived the final multivariable model using backward stepwise elimination by the Akaike Information Criterion. We externally validated the model in 3 international cohorts in Hong Kong (Queen Mary Hospital [HK-QMH], Ruttonjee Hospital [HK-RH]) and Japan (National Cerebral and Cardiovascular Center) by discrimination and calibration and compared its performance with the SeLECT and SeLECT2.0 scores. RESULTS: < 0.0001). For example, an IsCHEMiA score of 3 predicts a low risk of PSE at 1 year (2%) and 5 years (6%) while an IsCHEMiA score ≥8 predicts a high risk at 1 year (67%) and 5 years (78%). DISCUSSION: The IsCHEMiA score is an improved and readily applicable predictive model developed and validated using international stroke cohorts in the modern era of reperfusion therapies. It serves as a foundation for personalized management and may guide future clinical trials on antiepileptogenic therapies in acute ischemic stroke.