Скачать или смотреть Medical SAM3: A Foundation Model for Universal Prompt-Driven Medical Image Segmentation

Researchers have developed a new artificial intelligence tool called Medical SAM3 designed to automatically identify and outline specific body parts or diseases in medical images. While previous models worked well on regular photos, they often struggled with medical scans like X-rays, ultrasounds, and MRIs unless a human manually drew a box around the target first. To solve this, the team retrained an existing model using a massive collection of data from ten different types of medical imaging. This updated system can now accurately find and highlight complex details, such as tumors or blood vessels, simply by being told what to look for in text. Tests showed that Medical SAM3 performs much better than the original version, proving it can act as a reliable assistant for doctors by finding medical targets without needing extra manual hints.

https://arxiv.org/pdf/2601.10880
https://github.com/AIM-Research-Lab/M...