Enhancement of Stress ECG Performance with Machine Learning: A Single-Center Study.

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Exercise stress electrocardiogram (ECG) is a standard noninvasive tool for diagnosing coronary artery disease (CAD).
• The diagnostic accuracy of traditional exercise stress ECGs (ESEs) is often suboptimal, necessitating improved detection methods.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) model utilizing a transformer-based architecture.
• To enhance the diagnostic performance of exercise stress ECGs (ESEs) for detecting coronary artery disease (CAD).

Main Methods:

• An AI model was developed using a transformer architecture to process exercise stress ECG images into time-series data.
• The model integrated temporal ECG features to predict the presence of CAD.
• Performance was evaluated using 5-fold cross-validation on a test subset and subsequently on an independent validation cohort.

Main Results:

• The AI model achieved high diagnostic accuracy on the initial test subset, with 93.6% sensitivity, 93.2% specificity, and 93.4% overall accuracy.
• Significant improvements in sensitivity were observed, particularly in women (40.9% increase) and men (44.6% increase).
• In the second validation cohort, the model demonstrated 78% accuracy, 64.6% sensitivity, and 93% specificity.

Conclusions:

• This study provides proof of concept for an AI-based model in stress ECG interpretation.
• The developed AI model shows feasible and acceptable performance for enhancing CAD detection.