The abstract is read at the publisher; the summary is JClub's.
EchoLens, a transthoracic echocardiogram detector trained on 5,912 expert-annotated structures and 4,219 Doppler signals, recalled 87% and localized cardiac structures with 95.7% box-center accuracy.
- Why it matters: Current vision-language models lack spatial grounding, failing to identify the cardiac structures that support their echocardiography outputs. This limits interpretability and reliability of fine-grained structure localization.
- What they did: EchoLens was trained on data from 136 patients and covers 21 structures across 11 transthoracic views plus 11 spectral Doppler signal types. External evaluation by two physicians assessed detection correctness across all 11 views and Doppler signal type identification.
- The result: In external data, 81–92% of detections were judged correct across all 11 views, and the Doppler signal type was identified in 83% of frames. It enables spatial grounding for echocardiographic interpretation.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.