An ML-based Framework for Early Cerebral Stroke Prediction using Clinical Data

Cerebral stroke remains a major global health challenge, where early and accurate diagnosis is critical for reducing mortality and long-term disability. The challenges, such as missing data, class imbalance, and increasing complexity of clinical variables, limit the effectiveness of conventional diagnostic approaches. This study aims to develop an Artificial Intelligence (AI)-based framework for early, accurate, and explainable cerebral stroke prediction using structured clinical data.

This research proposes an ML-based framework that integrates clinical/tabular data for early prediction of cerebral stroke. The framework employed a Multi-layer Extra Trees Classifier (ETC) incorporating mean imputation and KNN imputation techniques, which are applied for handling missing values, while the Synthetic Minority Over-sampling technique (SMOTE) is used for class imbalance in clinical data. Experimental results demonstrate that the ETC achieves higher performance for early stroke risk assessment and improving decision-making in healthcare settings.

Experimental evaluation demonstrates that the proposed ETC combined with mean imputation and SMOTE significantly outperforms baseline models. The proposed model achieves an accuracy of 99.25%, precision of 98.85%, and recall of 98.92%, highlighting its robustness and effectiveness in early stroke prediction using clinical data.

The results indicate that appropriate data preprocessing techniques like mean imputation and SMOTE data balancing lead to substantial performance gains. The ETC shows predictive capability in capturing patterns in tabular clinical data. These findings support early stroke risk assessment and assist in decision-making in healthcare.

Overall, the proposed framework advances early cerebral stroke diagnosis using structured clinical data. By achieving high predictive performance, this work provides a strong foundation for AI-assisted stroke prediction using tabular healthcare data and contributes to preserving clinical decision support systems.