AI in Echocardiography
AI in Echocardiography
Source: JACC, 7 September 2025
1. Innovation
• Echo software model “EchoNet” with ability to automatically measure 18 echo parameters.
• Automates 18 key parameters (LV, RV, aorta, IVC, Doppler).
2. Data
• Trained on 150,000+ echo studies (Cedars-Sinai, 2011–2023).
• Validated on external datasets (Stanford and others).
• Largest dataset of sonographer annotations ever used.
3. Results
• High accuracy across both linear and Doppler measures.
• Reduces manual workload and increases reproducibility.
4. Limitations
• Still needs good-quality echo images — poor acquisition cannot be corrected by AI.
• Echo remains operator-dependent for image capture.
5. Impact
• Can support less-experienced users (e.g., ER doctors or nurses) for basic assessments, if images are adequate.
• Expands access, speeds up reporting, and standardizes measurements globally.
• AI software can run on older echo machines if they export digital images (DICOM); without digital output or good quality images, integration is not possible.
Take-home line:
AI can make echocardiography faster, more accurate, and more accessible — but image quality remains the key.