Researchers release Medical SAM3 for text-prompted segmentation across medical imaging modalities

Sign in✓ Signed in
Researchers led by Chongcong Jiang and Yu Tian have introduced Medical SAM3, a fully fine-tuned adaptation of SAM3 that performs universal, text-prompted medical image segmentation across 2D and 3D scans without privileged bounding boxes at test time. The work, posted on arXiv on 15 January 2026, argues that off-the-shelf SAM3 is a weak medical segmenter unless it is given strong geometric cues such as ground-truth boxes.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...