Enhancing Clinical Validation for Early Cardiovascular Disease Prediction through Simulation, AI, and Web Technology

Area of Science:

• Artificial Intelligence in Medicine
• Cardiovascular Disease Prediction
• Computational Health Science

Background:

• Cardiovascular diseases (CVDs) are a leading cause of mortality globally, necessitating advanced predictive models.
• Existing predictive models often lack the granularity to capture individual patient responses to risk factors.

Purpose of the Study:

• To develop an AI-driven dashboard for enhanced prediction of cardiovascular disease (CVD) progression.
• To improve the accuracy and user accessibility of CVD prediction models through agent-based simulation and ensemble learning.

Main Methods:

• Implemented an agent-based dynamic simulation technique to model individual patient responses to cardiovascular risk factors.
• Utilized ensemble learning, including XGBoost, achieving high accuracies (91-95%) in initial predictions.
• Integrated predictive models into a user-friendly Streamlit web application, achieving 97% predictive accuracy validated by Brier score and calibration curve.

Main Results:

• The AI dashboard achieved a 97% predictive accuracy for cardiovascular diseases (CVDs).
• Ensemble learning and XGBoost demonstrated strong performance with 91% and 95% accuracy, respectively.
• The Streamlit application facilitated seamless interaction for clinicians and patients, proving effective in real-time clinical settings.

Conclusions:

• Combining agent-based simulation, ensemble learning, and a user-centered web application significantly enhances CVD prediction accuracy and accessibility.
• The developed AI dashboard shows robust efficacy in early CVD detection and supports proactive patient management.
• Validated in an external clinical setting, the methodology demonstrates real-world applicability for improving cardiovascular health outcomes.