Enhancing cardiovascular risk prediction: the role of wall viscoelasticity in machine learning models

Area of Science:

• Biomedical Engineering
• Cardiovascular Research
• Machine Learning in Healthcare

Background:

• Cardiovascular disease (CVD) risk prediction relies on conventional factors like demographics and blood markers.
• Arterial wall mechanics, including viscoelasticity, offer potential for improved CVD risk stratification.
• Current models may not fully capture the complex biomechanical properties influencing CVD development.

Purpose of the Study:

• To evaluate the significance of arterial wall viscoelasticity in enhancing cardiovascular disease (CVD) risk prediction.
• To compare the impact of elastic versus viscous mechanical properties on predictive accuracy.
• To identify optimal machine learning models and features for CVD risk assessment.

Main Methods:

• Collected data on demographics, blood lab results, and arterial wall mechanical properties (Peterson's modulus, stiffness, energy dissipation ratio).
• Classified CVD risk (low/high) based on carotid ultrasound-derived total plaque area.
• Employed eight machine learning classifiers, including Random Forest Bagging Method (RFBM), and analyzed feature importance.

Main Results:

• Incorporating arterial wall mechanical attributes significantly improved predictive accuracies across most machine learning models.
• The RFBM achieved the highest performance (93.0% accuracy, 0.98 AUC) using all 10 features.
• Including viscous features (energy dissipation ratio) provided greater accuracy improvements than elastic features for tree-based bagging models.

Conclusions:

• Arterial wall viscoelasticity plays a crucial role in enhancing CVD risk prediction accuracy.
• Integrating viscous properties, such as energy dissipation ratio, offers significant advantages for predictive modeling.
• Combining both elastic and viscous wall characteristics holds potential for more precise CVD risk assessment.