EUCARDIA: a web-based platform for the CVD prediction using ML techniques in the Greek population

Area of Science:

• Medical Informatics
• Cardiovascular Research
• Machine Learning Applications

Background:

• Traditional statistical models for cardiovascular disease (CVD) risk estimation have limitations, often leading to inaccurate predictions.
• Accurate CVD risk assessment is crucial for timely intervention and prevention strategies.

Purpose of the Study:

• To develop and implement a web-based Machine Learning (ML) platform for personalized 10-year CVD risk prediction.
• To address the methodological constraints of existing statistical models for the Greek population.

Main Methods:

• A retrospective study utilized clinical and demographic data from 3,290 CVD-free participants.
• Two ML classifiers were developed: a binary classifier for CVD occurrence and a multiclass classifier for SCORE2 risk stratification.
• Algorithm selection was based on ROC-AUC evaluation, with Voting Ensemble and Stacking Ensemble chosen for binary and multiclass tasks, respectively.

Main Results:

• The binary classifier achieved an AUC-ROC of 0.78, and the multiclass classifier achieved 0.97.
• The ML platform demonstrated superior performance compared to the traditional HellenicScore model across most metrics.
• A web-based platform integrating the ML model and scientific literature retrieval was successfully deployed.

Conclusions:

• The developed ML platform serves as a reliable tool for CVD risk assessment.
• The platform shows potential to outperform traditional statistical models in predicting personalized CVD risk.