Magnetocardiography-based coronary artery disease severity assessment and localization using spatiotemporal features

Area of Science:

• Biomedical Engineering
• Cardiology
• Medical Imaging

Background:

• Coronary artery disease (CAD) diagnosis relies on invasive or less specific imaging methods.
• Magnetocardiography (MCG) offers a non-invasive alternative for cardiac electrical activity assessment.
• Automated analysis of MCG data for CAD is an unmet clinical need.

Purpose of the Study:

• To develop an automated and accurate method for assessing CAD severity and localization using optically pumped magnetometer MCG.
• To identify key spatiotemporal features from MCG signals for CAD assessment.
• To classify CAD severity (absent, mild, moderate, severe) and pinpoint stenosis location (LAD, LCX, RCA).

Main Methods:

• Utilized spatiotemporal features from MCG signals: amplitude, correlation, local binary pattern, and shape.
• Employed machine learning models, including Support Vector Machine (SVM) and discriminant analysis.
• Trained models to classify CAD severity and artery-specific locations (LAD, LCX, RCA).

Main Results:

• SVM achieved 75.1% accuracy for CAD severity assessment (AUC 0.876).
• Highest accuracy for localization: 94.3% (LAD) via SVM, 84.4% (LCX) and 84.9% (RCA) via discriminant analysis.
• Amplitude and correlation features were identified as crucial for both severity and localization.

Conclusions:

• The developed automated method effectively assesses CAD severity and localizes blockages using MCG.
• Machine learning models provide accurate and automated diagnostic capabilities for CAD interpretation.
• This approach has the potential to enhance clinical acceptance and application of MCG in cardiology.