Enhancing Stroke Prediction through Interpretable AI: Distinguishing Stroke Cases from Non-Stroke Cases

A sudden interruption of blood supply to part of the brain can result in a stroke, causing impairment and possible fatalities due to damage to brain cells. Rapid identification and intervention are crucial for preventing strokes and minimizing severe brain damage. The integration of Artificial Intelligence (AI) and machine learning (ML) shows promise in improving healthcare practitioners' ability to predict strokes swiftly and accurately. This study focuses on creating and assessing a stroke prediction model using a Random Forest (RF) classifier, supported by various performance metrics like AUC, precision, recall, F1-score, and accuracy, achieving a commendable accuracy rate of 94.6%. To address the class imbalance in the training set, a Random Over-Sampler was employed. Moreover, interpretability is enhanced through Explainable AI (XAI) techniques, specifically Shapley Additive Values (SHAP) and Local Interpretable Modelagnostic Explanations (LIME). By establishing an automated screening framework, this research aims to provide healthcare professionals with tools to offer more personalized and efficient care, potentially transforming stroke prevention and treatment.