Stroke Risk Factor Prediction Using Machine Learning Techniques: A Systematic Review

Olusola Olabanjo, Ashiribo Wusu, Oseni Afisi and Boluwaji Akinnuwesi

This review addresses the global challenge of stroke, a leading cause of disability and mortality. The unpredictability and severe impact of stroke necessitate advanced prediction methods. In this work, the machine learning (ML) and deep learning (DL) techniques in stroke risk prediction were evaluated, assessing their effectiveness and application in diverse contexts. A systematic analysis of existing studies and datasets was conducted using Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA), focusing on various ML and DL algorithms used in stroke risk prediction. The 31 papers met the final inclusion criteria. The review highlights significant advancements in stroke prediction using ML and DL models, noting their ability to manage complex datasets and provide accurate predictions. However, challenges such as the need for external validation, model explainability and model transparency persist. Feature importance is further recommended to offer context-specific recommendations as stroke risk factors vary in different countries. This study also spotlights Random Forest as the outperforming model in predicting stroke risks, secondary data as the prominent dataset and China, India and Bangladesh as the country with the most stroke risk studies. The ML and DL offer promising tools for stroke risk prediction, enhancing personalized healthcare strategies. Addressing existing challenges will be crucial for their effective integration into clinical practice.

View Fulltext Back