OrthoDiffusion: A Generalizable Multi-Task Diffusion Foundation Model for Musculoskeletal MRI Interpretation
Tian Lan, Lei Xu, Zimu Yuan, Shanggui Liu, Jiajun Liu, Jiaxin Liu, Weilai Xiang, Hongyu Yang, Dong Jiang, Jianxin Yin, Dingyu Wang · 2026
Read OrthoDiffusion: A Generalizable Multi-Task Diffusion Foundation Model for Musculoskeletal MRI Interpretation by Tian Lan, Lei Xu, Zimu Yuan, Shanggui Liu, Jiajun Liu, Jiaxin Liu, Weilai Xiang, Hongyu Yang, Dong Jiang, Jianxin Yin, Dingyu Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20752.txt.
Description
Muscle-bone ills are a large global health burden and a leading cause of harm, yet MRI reading remains challenging. Tian Lan and ten co-authors built OrthoDiffusion, a unified diffusion-based model for multi-task MRI reading.
The system uses three view 3D models, pre-trained in a self-taught way on nearly sixteen thousand unmarked knee MRI scans. It learns body-part features from side, front, and top views. Those view features support tasks, including body-part mapping and multi-label calls. Tests show it achieves results segmenting eleven knee structures and detecting eight knee flaws. The model shows strength across centers and field levels, always topping older models. Where tagged data was rare, it kept high test hits using only ten percent of labels. Features learned from knee imaging proved mobile to other joints, with diagnostic results across eleven ills of the ankle and shoulder.
Catalog record
| CREATOR | Tian Lan, Lei Xu, Zimu Yuan, Shanggui Liu, Jiajun Liu, Jiaxin Liu, Weilai Xiang, Hongyu Yang, Dong Jiang, Jianxin Yin, Dingyu Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 83 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20752