Kev-Bio vision-language model for ultrasound biometry
Original title: Every Pixel Is an Option: A System One Model for Medical Ultrasound Biometry
Kev-Bio uses language model decisions to generate queries for landmark detection on the FU/GU Biometry 2026 dataset.
The paper proposes Kev-Bio, a vision-language model inspired by System One decision models. It guides ultrasound landmark detection using clinical measurement questions and produces spatial probability maps for coordinate prediction. On the nine-view FU/GU Biometry 2026 dataset, Kev-Bio achieves a mean radial error of 22.50 pixels and a clinical measurement MAE of 15.38, outperforming baselines. Code and models are available on GitHub.
- Mean radial error of 22.50 pixels on nine-view dataset
- Clinical measurement MAE of 15.38
- Trained end-to-end with landmark and measurement losses
- Uses shared dense ultrasound features for probability maps
From the original article
1 min read at huggingface.co
**Purpose:** Ultrasound biometry is widely used across various clinical applications, including prenatal and cardiac examinations, to obtain quantitative measurements by placing calipers at anatomical landmarks. However, these measurements remain subject to interoperator variability as they rely largely on manual caliper placement.
By Moein Heidari, Junbo Rao, Yudan Chen · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.