EchoNet-Measurements is a novel, open-source deep learning model that can accurately quantify 18 anatomic and Doppler measurements in echocardiography. Developed by researchers at Cedars-Sinai Medical Center, this AI model shows promise in automating the process of taking measurements, which could save time and increase accuracy.
The model directly annotates the areas that cardiologists would have annotated themselves, making the results more interpretable and robust. EchoNet-Measurements demonstrated high levels of accuracy across all measurements compared to sonographer measurements and an external validation data set. The algorithm's performance was consistent across patient characteristics, including age, sex, atrial fibrillation, and obesity status, as well as across machine vendors.
If implemented clinically, EchoNet-Measurements could save up to 10-20 minutes of time per study, increasing efficiency and productivity. The model could also potentially provide more precise measurements than human sonographers, reducing variability and errors.
As a free and open-source tool, EchoNet-Measurements serves as a valuable benchmark for clinicians to evaluate AI algorithms. While commercial products like (link unavailable) system exist, the model's developers hope it will push the field to develop better algorithms with more training data.
The researchers are planning a prospective trial to compare the performance of EchoNet-Measurements against human sonographers. This trial will help determine the model's effectiveness in real-world clinical settings and its potential to improve patient care.
