Explainable AI Modeling in the Prediction of Cardiovascular Disease Risk.

Area of Science:

• Artificial Intelligence in Medicine
• Machine Learning for Healthcare
• Cardiovascular Disease Research

Background:

• Cardiovascular disease (CVD) prediction models often lack transparency.
• Explainable AI (XAI) is crucial for clinical trust and adoption.
• Current machine learning models may not provide sufficient interpretability.

Purpose of the Study:

• To develop an explainable AI (XAI) model for cardiovascular disease (CVD) prediction.
• To integrate rule extraction and argumentation theory for enhanced model interpretability.
• To improve accuracy and provide both global and local explanations for CVD risk.

Main Methods:

• Utilized the XGBoost machine learning algorithm for CVD risk prediction.
• Applied rule extraction techniques to the XGBoost model.
• Incorporated argumentation theory to provide interpretability and explainability.
• Evaluated model performance in scenarios with low confidence results or dilemmas.

Main Results:

• The developed XAI model demonstrated agreement with previous XGBoost findings for CVD risk prediction.
• Rule-based explainability was successfully integrated, offering significant advantages.
• The model provided both global and local explainability for CVD risk factors.
• The approach showed potential in handling low-confidence predictions and decision dilemmas.

Conclusions:

• The study successfully developed an explainable AI model for cardiovascular disease prediction.
• The integration of XGBoost with rule extraction and argumentation theory enhances model transparency.
• Further research is needed to refine argumentation-based interpretability for complex clinical scenarios.
• The findings support the use of XAI in improving the reliability and clinical utility of CVD risk prediction models.