Related Experiment Video
Updated: Sep 16, 2025
06:37
Electroencephalography Network Indices as Biomarkers of Upper Limb Impairment in Chronic Stroke
Published on: July 14, 2023
977
Machine learning techniques for stroke prediction: A systematic review of algorithms, datasets, and regional gaps
Afeez Adekunle Soladoye1, Nicholas Aderinto2, Mayowa Racheal Popoola3
1Department of Computer Engineering, Federal University Oye-Ekiti, Ekiti, Nigeria.
International Journal of Medical Informatics
|July 12, 2025
Summary
Machine learning (ML) shows promise for stroke prediction, but high-risk populations are underrepresented. Future research needs to focus on developing and validating ML models for diverse patient groups to improve clinical utility.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
Background:
- Stroke is a major global cause of death and disability, affecting 15 million people annually.
- Machine learning (ML) offers data-driven approaches for early stroke risk identification.
- Systematic evaluation of ML's clinical utility in stroke prediction is needed.
Purpose of the Study:
- To systematically review ML techniques for stroke prediction.
- To synthesize performance metrics, clinical applicability, and research trends.
- To analyze patient demographics and stroke prevalence patterns in ML studies.
Main Methods:
- Systematic review adhering to PRISMA guidelines.
- Searched five databases for open-access ML-based stroke prediction studies (2013-2024).
- Extracted data on methodology, datasets, metrics, targets, sources, demographics, and prevalence; descriptive synthesis used.
Main Results:
- 58 studies included, with peak publications in 2021.
- Primary objectives: stroke occurrence (62.7%), outcome prediction (22.9%), and type classification (14.4%).
- Electronic health records (57.8%) and imaging (25.3%) were key data sources; ensemble methods (90.4-97.8%) and deep learning showed high accuracy. African populations were underrepresented.
Conclusions:
- ML techniques demonstrate potential for stroke prediction.
- Significant gaps exist in high-risk population representation and clinical validation.
- Future research should focus on population-specific models and clinical implementation.
