Prediction and critical feature analysis for coronary artery calcification progression

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Coronary calcification is a significant cardiovascular risk factor, but its progression is not well understood.
• Predicting coronary artery calcification progression is challenging due to limited clinical data and the complexity of traditional risk models.
• Accurate prediction is crucial for effective risk stratification and individualized treatment strategies in patients with coronary artery disease.

Purpose of the Study:

• To evaluate the efficacy of traditional machine learning models in predicting coronary artery calcification progression using serial CCTA data.
• To identify key predictors of coronary artery calcification progression from CCTA features.
• To develop and validate a predictive score for personalized risk assessment of coronary artery calcification progression.

Main Methods:

• A dataset of 2,579 patients with serial CCTA scans was analyzed.
• Machine learning models including Random Forest, Gradient Boosting Decision Trees, XGBoost, and Logistic Regression were employed.
• SHAP analysis was used to identify key predictive features and interpret model outputs.

Main Results:

• The Random Forest model demonstrated superior predictive performance (AUC = 0.81) compared to traditional clinical models (AUC = 0.64).
• Key predictors identified included baseline coronary artery calcium score (CACS), plaque burden, and other CCTA-derived features.
• A coronary artery calcification progression prediction score (CACPPS) was developed, showing good calibration and clinical utility.

Conclusions:

• Interpretable machine learning models can accurately predict coronary artery calcification progression from CCTA.
• The developed CACPPS provides a tool for dynamic, individualized risk management in patients with suspected or confirmed coronary artery disease.
• This approach addresses 'black-box' concerns and supports clinical decision-making for better patient outcomes.