Electrocardiogram-Based Artificial Intelligence to Identify Coronary Artery Disease

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Coronary artery disease (CAD) is a major cause of illness and death.
• Early detection of CAD is crucial for effective management and prevention of adverse outcomes.

Purpose of the Study:

• To develop a deep learning model (ECG2CAD) for detecting CAD using electrocardiograms (ECGs).
• To identify individuals at risk for adverse events in a primary care setting using ECG data.

Main Methods:

• Trained ECG2CAD on over 760,000 ECGs from Massachusetts General Hospital (MGH).
• Evaluated model performance using AUROC and AUPRC in MGH, Brigham and Women's Hospital (BWH), and UK Biobank datasets.
• Assessed risk for incident CAD-related events in a BWH primary care cohort.

Main Results:

• ECG2CAD demonstrated consistent performance in discriminating prevalent CAD across all test sets.
• The model showed incremental value over traditional risk models (age, sex, Pooled Cohort Equations).
• High ECG2CAD risk scores correlated with significantly increased risk of myocardial infarction, heart failure, and all-cause mortality.

Conclusions:

• AI-powered ECG analysis can aid in identifying individuals with potential undiagnosed CAD.
• This technology can inform further diagnostic testing and preventive strategies for CAD.
• ECG2CAD offers a promising tool for risk stratification in primary care settings.