Prediction of post-PCI angina risk using machine learning-based magnetocardiography model

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Recurrent angina post-percutaneous coronary intervention (PCI) significantly impacts patient quality of life.
• The predictive capability of noninvasive Magnetocardiography (MCG) for symptomatic outcomes after PCI is not well-defined.

Purpose of the Study:

• To develop and validate a machine learning-based MCG model for predicting angina after PCI.
• To assess the added value of combining MCG with clinical biomarkers for enhanced angina prediction.

Main Methods:

• 110 patients undergoing successful PCI were enrolled, with MCG performed pre- and post-procedure.
• Angina status was assessed using Seattle Angina Questionnaire (SAQ) domains (SAQ-AS and SAQ-AF).
• Multivariable logistic regression was used to build MCG and combined models, evaluated by AUC, sensitivity, and F1-score.

Main Results:

• The combined model integrating MCG and biomarkers showed superior discriminatory ability for angina stability (SAQ-AS).
• For angina frequency (SAQ-AF), the combined model outperformed MCG alone with higher sensitivity (0.813 vs. 0.801) and AUC (0.813 vs. 0.801).
• Nomograms provided effective risk stratification, and model calibration was satisfactory.

Conclusions:

• A machine learning-based MCG model offers an effective, noninvasive method for predicting post-PCI angina risk.
• Integrating MCG with clinical biomarkers enhances risk stratification, aiding in the identification of high-risk patients.