Explainable machine learning for predicting coronary heart disease risk in patients with carotid atherosclerosis: A...

Area of Science:

• Cardiovascular Medicine
• Biomedical Data Science
• Machine Learning in Healthcare

Background:

• Carotid atherosclerosis significantly elevates coronary heart disease (CHD) risk.
• Existing risk prediction models for CHD in this population lack specificity and interpretability.
• There is a need for explainable models to improve risk assessment in patients with carotid atherosclerosis.

Purpose of the Study:

• To develop and validate explainable machine learning (ML) models for predicting CHD risk.
• To identify key predictors of CHD in patients with carotid atherosclerosis.
• To enhance the interpretability and clinical utility of CHD risk prediction models.

Main Methods:

• Retrospective analysis of 487 patients with carotid atherosclerosis.
• Feature selection using LASSO regression to identify six key predictors.
• Training and evaluation of seven ML models, including logistic regression, using AUC, PRC-AUC, calibration curves, and Decision Curve Analysis (DCA).
• Application of SHAP (SHapley Additive exPlanations) for model interpretability.

Main Results:

• The logistic regression model demonstrated superior performance with an AUC of 0.827 and PRC-AUC of 0.752.
• SHAP analysis highlighted age and diastolic blood pressure as the most influential predictors.
• The logistic regression model showed strong generalizability, calibration, and superior clinical net benefit via DCA.

Conclusions:

• A six-variable logistic regression model offers accurate and interpretable CHD risk prediction for patients with carotid atherosclerosis.
• The model's transparency and clinical utility support its application in personalized risk management strategies.
• Explainable ML models can significantly improve cardiovascular risk assessment and patient care.