Magnetocardiography-Based Ischemic Heart Disease Detection and Localization Using Machine Learning Methods.

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Ischemic heart disease (IHD) diagnosis relies on complex interpretations of cardiac signals.
• Magnetocardiography (MCG) offers a non-invasive method to assess cardiac magnetic fields.
• Developing automated diagnostic tools for IHD is crucial for timely clinical intervention.

Purpose of the Study:

• To develop a rapid and precise automated methodology for detecting and localizing ischemic heart disease.
• To evaluate the efficacy of machine learning classifiers in analyzing MCG data for IHD diagnosis.
• To explore the correlation between MCG features and the location of coronary artery stenosis.

Main Methods:

• Extracted 164 features from T waves in averaged MCG recordings, categorized into time, frequency, and information theory domains.
• Compared k-nearest neighbor, decision tree, SVM, and XGBoost classifiers for IHD detection, employing model ensemble for best performers.
• Utilized XGBoost with 18 time-domain features for localizing ischemia to specific coronary arteries (LAD, LCX, RCA).

Main Results:

• The SVM-XGBoost ensemble model achieved 94.03% accuracy for IHD detection, with high precision and recall.
• For ischemia localization, XGBoost demonstrated accuracies of 74% (LAD), 68% (LCX), and 65% (RCA).
• T wave repolarization synchronicity and magnetic field patterns were identified as key indicators for IHD and stenosis location.

Conclusions:

• An automated system for IHD detection and localization using MCG and machine learning has been successfully developed.
• T wave repolarization synchronicity is a significant factor in differentiating IHD from normal subjects.
• MCG magnetic field patterns correlate with ischemia location, enabling non-invasive localization.