Explainable AI-Driven Decision Support System for Early Prediction of Cardiovascular Diseases

Abstract

Cardiovascular diseases (CVDs) remain a major cause of morbidity and mortality worldwide, emphasizing the need for reliable methods that can identify high-risk individuals at an early stage. Although machine learning and deep learning approaches have demonstrated considerable potential for cardiovascular risk prediction, their limited interpretability often restricts their acceptance in clinical decision-making. This study proposes an Explainable AI-Driven Decision Support System (XAI-DSS) for the early prediction of cardiovascular diseases by integrating intelligent clinical data preprocessing, feature selection, an ensemble learning-based prediction model, and explainable artificial intelligence. The framework processes heterogeneous cardiovascular risk factors, including demographic characteristics, blood pressure, cholesterol, glucose levels, electrocardiographic attributes, lifestyle factors, and other relevant clinical indicators. A hybrid feature-selection strategy is employed to identify the most informative risk variables, while an optimized ensemble classifier generates patient-specific CVD risk predictions. Explainability is incorporated using SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) to provide both global and patient-level interpretations of model decisions. Experimental evaluation demonstrated an accuracy of 96.42%, sensitivity of 95.81%, specificity of 96.87%, precision of 96.15%, F1-score of 95.98%, and area under the ROC curve (AUC) of 0.982. Compared with the selected baseline machine-learning model, the proposed XAI-DSS achieved an approximately 4.7% improvement in prediction accuracy and 5.3% improvement in F1-score. Explainability analysis further identified age, systolic blood pressure, cholesterol, maximum heart rate, fasting blood glucose, and chest-pain characteristics as influential factors contributing to cardiovascular risk predictions. The proposed framework therefore combines high predictive performance with transparent clinical reasoning, enabling healthcare professionals to understand the factors influencing individual risk assessments. The developed XAI-DSS can serve as a supportive screening framework for early cardiovascular risk stratification and informed clinical decision-making, subject to external clinical validation.