Artificial Intelligence Methods in Stroke Prediction: A Narrative Review

Abstract

Stroke is a leading cause of mortality worldwide. Despite rising healthcare costs, early noninvasive stroke risk stratification via artificial intelligence provides accurate, rapid predictions, establishing AI as a powerful tool in healthcare by enabling tailored treatment for stroke patients. Traditional algorithms assess risk using regression-based statistical correlations, which, while practical and user-friendly, show moderate predictive accuracy. Machine learning (ML) and deep learning (DL) models effectively identify individuals at higher risk of stroke across diverse populations, enabling timely intervention. However, evidence summarizing the effectiveness of these approaches remains sparse. This review presents various ML and DL algorithms, clarifying their theoretical foundations, historical evolution, and applications in stroke prediction. This narrative review focuses on studies published within the last five years and emphasizes the clinical relevance and translational potential of AI-based stroke prediction models.