Machine Learning Model for Atherosclerosis Evaluation and Cardiovascular Risk Prediction Based on Coronary CT...

Area of Science:

• Cardiovascular Imaging
• Machine Learning in Medicine
• Predictive Analytics

Background:

• Traditional risk prediction tools for atherosclerotic cardiovascular disease (ASCVD) have limitations.
• Coronary artery calcium (CAC) scoring is a common risk assessment tool but has limitations.

Purpose of the Study:

• To develop and validate a machine learning model using coronary computed tomography angiography (CCTA) data for improved ASCVD risk prediction.
• To compare the performance of the CCTA-based machine learning model against traditional risk factors and CAC scoring.

Main Methods:

• The CREATION study included 8431 patients with suspected coronary artery disease undergoing CCTA.
• Six machine learning survival models were trained using 48 CCTA parameters; XGBoost was selected for model development.
• The primary outcome was major adverse cardiac events (MACE): all-cause death, myocardial infarction, revascularization, or stroke.

Main Results:

• The XGBoost CCTA model significantly outperformed traditional risk factors and CAC scoring in both training (AUC 0.903 vs 0.830) and testing (AUC 0.899 vs 0.753) cohorts.
• Key predictors included diameter stenosis, lipid plaque burden, total plaque volume, high-risk plaque, and vessel volume.
• Lipid plaque burden showed a strong association with MACE (aHR 2.524 per 5% increase).

Conclusions:

• A machine learning model integrating CCTA plaque quantification, characterization, and stenosis assessment significantly enhances MACE prediction.
• This CCTA-based model provides direct visualization of coronary atherosclerosis and outperforms current clinical practice models.
• The findings highlight the potential of advanced CCTA analysis for more accurate cardiovascular risk stratification.