Related Experiment Videos
Predictive Performance of Artificial Intelligence Models for Stroke Risk Stratification: A Systematic Review and
1Social Determinants of Health Research Center, Semnan University of Medical Sciences, Semnan, Iran.
Summary
Artificial intelligence (AI) models show promise in predicting stroke risk, with deep learning (DL) models performing well. However, more standardized research and validation are needed for widespread clinical use.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
- Public Health
Background:
- Stroke is a major global cause of death and disability, straining healthcare systems.
- Accurate stroke risk stratification is crucial for targeted prevention and resource optimization.
- Existing research on AI for stroke risk prediction needs comprehensive synthesis.
Purpose of the Study:
- To systematically evaluate the predictive performance of AI models for stroke risk assessment.
- To synthesize evidence through meta-analysis on AI-based stroke risk stratification.
- To explore the implications of AI in stroke risk prediction for healthcare planning.
Main Methods:
- Systematic literature search across major databases (Web of Science, PubMed, Scopus) until January 2025.
- Adherence to PRISMA 2020 guidelines for systematic reviews.
- Meta-analysis of Area Under the Receiver Operating Characteristic Curve (AUC) values for different AI algorithms.
Main Results:
- Deep learning (DL) models exhibited strong discriminative performance with a pooled AUC of 0.955.
- Imaging-based DL models showed particularly favorable predictive capabilities.
- Significant heterogeneity was observed across studies, impacting generalizability.
Conclusions:
- AI, especially DL, demonstrates high potential for stroke risk stratification.
- Methodological heterogeneity and limited external validation hinder immediate clinical implementation.
- Standardized reporting and validation frameworks (e.g., TRIPOD) are essential for future AI research in stroke prediction.