A Powerful Paradigm for Cardiovascular Risk Stratification Using Multiclass, Multi-Label, and Ensemble-Based Machine...

Area of Science:

• Cardiology
• Artificial Intelligence
• Machine Learning

Background:

• Cardiovascular disease (CVD) is a leading global cause of mortality.
• Escalating healthcare costs necessitate early, non-invasive CVD risk assessment.
• Artificial Intelligence (AI) methods offer superior performance over conventional approaches for CVD risk prediction.

Purpose of the Study:

• To review the three most recent AI paradigms for CVD risk assessment: multiclass, multi-label, and ensemble-based methods.
• To analyze these methods in both office-based and stress-test laboratory settings.
• To evaluate AI risk-of-bias (RoB) within the CVD framework.

Main Methods:

• Systematic review of 265 CVD studies using PRISMA guidelines.
• Analysis of multiclass, multi-label, and ensemble AI methods using machine learning (ML) frameworks.
• Comprehensive review of AI methods based on architecture, applications, pros/cons, validation, and RoB.

Main Results:

• Office-based, laboratory-based, image-based phenotypes, and medication usage are common biomarkers.
• Surrogate carotid scanning shows promise for coronary artery risk prediction.
• Multiclass classification is the most popular paradigm, followed by ensemble and multi-label methods.

Conclusions:

• AI-based methods are highly promising for CVD risk assessment.
• Accurate ground truth (GT) selection is vital to prevent RoB in AI models.
• Combining image-based strategies with conventional risk factors enhances stability in AI CVD risk assessment frameworks.