A multimodal knowledge-enhanced whole-slide pathology foundation model
Yingxue Xu, Yihui Wang, Fengtao Zhou, Jiabo Ma, Cheng Jin, Shu Yang, Jinbang Li, Zhengyu Zhang, Chenglong Zhao, Huajun Zhou, Zhenhui Li, Huangjing Lin, Xin Wang, Jiguang Wang, Anjia Han, Ronald Cheong Kin Chan, Li Liang, Xiuming Zhang, Hao Chen · 2025
Read A multimodal knowledge-enhanced whole-slide pathology foundation model by Yingxue Xu, Yihui Wang, Fengtao Zhou, Jiabo Ma, Cheng Jin, Shu Yang, Jinbang Li, Zhengyu Zhang, Chenglong Zhao, Huajun Zhou, Zhenhui Li, Huangjing Lin, Xin Wang, Jiguang Wang, Anjia Han, Ronald Cheong Kin Chan, Li Liang, Xiuming Zhang, Hao Chen (2025) 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/pmc-12738713.txt.
Description
Yingxue Xu and co-authors built a pathology foundation model that learns from slides, expert reports, and gene data together, in a study published in 2025. Most models read only images or image captions, and zoom into small patches, missing the whole slide. Their model, mSTAR, trains on over twenty-six thousand slide-level bundles across thirty-two cancer types, holding more than one hundred sixteen million patch images. Whole-slide context from all three sources is injected into each patch's representation.
Across ninety-seven oncology benchmark tasks, mSTAR beat prior state-of-the-art models. The biggest gains came in molecular prediction and tasks mixing data types. The authors conclude that fusing modalities teaches more than simply scaling image data. They offer the model as a base for slide-level cancer analysis that sees tissue, text, and genes at once.
Catalog record
| CREATOR | Yingxue Xu, Yihui Wang, Fengtao Zhou, Jiabo Ma, Cheng Jin, Shu Yang, Jinbang Li, Zhengyu Zhang, Chenglong Zhao, Huajun Zhou, Zhenhui Li, Huangjing Lin, Xin Wang, Jiguang Wang, Anjia Han, Ronald Cheong Kin Chan, Li Liang, Xiuming Zhang, Hao Chen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 113 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |