A machine learning model using echocardiographic myocardial strain to detect myocardial ischemia
Bo Zheng1,2, Yaokun Liu1, Jingyi Zhang3
1Department of Cardiology, Peking University First Hospital, Beijing, China.
Internal and Emergency Medicine
|May 21, 2025
Summary
Artificial intelligence (AI) and machine learning can now detect myocardial ischemia using echocardiographic strain, offering a non-invasive alternative. This AI model achieved 85.9% accuracy, showing promise for clinical use.
Area of Science:
- Cardiology
- Medical Imaging
- Artificial Intelligence
Background:
- Coronary functional assessment is vital for revascularization decisions.
- Traditional ischemia detection methods have limitations.
- Echocardiographic myocardial strain offers a non-invasive approach but faces inter-operator variability.
Purpose of the Study:
- To develop a non-invasive, AI-driven solution for automated myocardial ischemia detection.
- To reduce variability in strain-based ischemia assessment.
- To create a robust diagnostic tool for clinical practice.
Main Methods:
- An AI model was trained using echocardiographic myocardial strain data and six clinical features.
- Left ventricular endocardium tracing extracted strain data.
- Coronary angiography-derived fractional flow reserve (caFFR) ≤ 0.80 defined myocardial ischemia.
- Ensemble-learning algorithms were used for model training and optimization.
Main Results:
- The study included 636 subjects with suspected coronary artery disease; 44.3% had myocardial ischemia.
- The AI model achieved 85.9% diagnostic accuracy, 88.9% sensitivity, and 83.1% specificity.
- The model's area under the receiver operating characteristic curve (AUC) was 0.915.
Conclusions:
- The AI prototype model based on echocardiographic myocardial strain shows promising results for detecting myocardial ischemia.
- This non-invasive method has the potential to improve clinical decision-making.
- Further validation on larger patient populations is required.
