Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
Ultrasonography is an imaging technique that uses the reflections of high-frequency sound waves to create an image of a structure located within the body.
Holste et al. present a deep learning method for efficient cardiac disease diagnosis from echocardiography videos. Their method, EchoCLR, uses self-supervised learning to accurately detect heart diseases from a smaller number of labeled examples than existing methods.
Vision–language models can be trained to read cardiac ultrasound images with implications for improving clinical workflows, but additional development and validation will be required before such models can replace humans.
New findings provide insight into the natural history of subclinical synovitis, a reported predictor of the development of rheumatoid arthritis, and identify various factors associated with its reversal.
An artificial intelligence-guided workflow for initial evaluation of left ventricular ejection fraction in echocardiography is non-inferior to initial assessment by a sonographer, according to findings from a blinded, randomized, non-inferiority clinical trial.
Researchers have engineered a wearable device that adheres to the skin and uses ultrasound imaging and a deep learning model to produce a dynamic, real-time assessment of cardiac function.
New advances in machine learning could facilitate and reduce disparities in the prenatal diagnosis of congenital health disease, the most common and lethal birth defect.