A new open-source artificial intelligence (AI) model can automate the measurement of key echocardiographic parameters with high accuracy, potentially reducing clinician workload and improving consistency, according to a new study published in JACC.¹
The model, named EchoNet-Measurements, uses deep learning semantic segmentation to automatically analyse and quantify 18 different anatomic and Doppler measurements from transthoracic echocardiography (TTE) images. By automating this time-consuming manual task, the tool aims to streamline the echocardiography workflow.
Researchers developed and validated the model using a large dataset from the Cedars-Sinai Medical Center (CSMC), Los Angeles, US, comprising 877,983 echocardiographic measurements from 155,215 studies conducted between 2011 and 2023. The model was trained to assess nine B-mode and nine Doppler measurements.
The performance of EchoNet-Measurements was compared against manual measurements by sonographers in a held-out internal test set from CSMC and an independent external validation cohort from Stanford Healthcare (SHC), Palo Alto, US. A further end-to-end evaluation was performed on 2,103 temporally distinct studies at CSMC to assess performance in a simulated clinical workflow.
The AI model demonstrated high accuracy and strong agreement with sonographer measurements across both internal and external validation cohorts. The mean coverage probability (CP)—the likelihood that two measurements differ by less than a predefined acceptable amount—was 0.796 in the CSMC held-out test set and 0.839 in the external SHC dataset.
In the end-to-end analysis, the model achieved a mean CP of 0.803 and a mean relative difference of 0.108. The performance remained consistent across various patient characteristics, including age, sex, atrial fibrillation, and obesity status, as well as across different ultrasound machine vendors.
These findings suggest that AI can reliably automate a significant portion of the echocardiographic measurement process, which could lead to increased efficiency and reduced inter-observer variability in clinical practice. The study authors concluded that, “EchoNet-Measurements achieves high accuracy in automated echocardiographic quantification and potential for assisting the clinicians in the echocardiography workflow. This open-source model provides the foundation for future developments in artificial intelligence applied to echocardiography.”¹
The authors noted that the current model does not yet include all parameters assessed in clinical echocardiography, such as velocity time integral and left atrium volume. Future studies are warranted to develop these additional measurements, and prospective randomised trials will be necessary before widespread clinical deployment.
This study was funded by the National Institutes of Health and the National Heart, Lung, and Blood Institute.
References
1. Sahashi Y, Ieki H, Yuan V, et al. Artificial Intelligence Automation of Echocardiographic Measurements. JACC. 2025;86(13):964-978. https://doi.org/10.1016/j.jacc.2025.07.053
2. Tromp J, Seekings PJ, Hung CL, et al. Automated interpretation of systolic and diastolic function on the echocardiogram: a multicohort study. Lancet Digit Health. 2022;4(1):e46-e54. https://doi.org/10.1016/S2589-7500(21)00235-1
Disclaimer: The information presented in this article is for educational purposes only. Any quotes included reflect the opinions of the individual quoted, and do not necessarily reflect the views of the publisher. The publisher does not guarantee the accuracy or completeness of the content and accepts no responsibility for any errors, or any consequences arising from its use.