bifold-pathomics/PathoROB-tcga

PathoROB

Preprint | Code | Licenses | Cite

PathoROB is a benchmark for the robustness of pathology foundation models (FMs) to non-biological medical center differences.

PathoROB contains four datasets covering 28 biological classes from 34 medical centers and three metrics:

1. Robustness Index: Measures the dominance of biological over non-biological features in an FM representation space.
2. Average Performance Drop (APD): Measures the robustness of downstream models to shortcut learning of non-biological features and the effect on generalization performance.
3. Clustering Score: Measures the robustness of clustering to non-biological features and the impact on the quality of k-means clusters.

How to use

Check out our PathoROB GitHub repository.

Latest updates

• December 2025: PathoROB codes are available on GitHub.
• September 2025: PathoROB data are available on Hugging Face.

Licenses

The PathoROB datasets were subsampled from public sources. Therefore, we redistribute each PathoROB dataset under the license of its original data source. You can run PathoROB on any subset of datasets with licenses suitable for your application.

• Camelyon:
  • Source: CAMELYON16 and CAMELYON17
  • License: CC0 1.0 (Public Domain)
• TCGA:
  • Source: TCGA-UT
  • License: CC-BY-NC-SA 4.0 (Non-Commercial Use)
• Tolkach ESCA
  • Source: https://zenodo.org/records/7548828
  • License: CC-BY-SA 4.0
  • Comment: This license was granted by the author specifically for PathoROB.

Acknowledgements

We want to thank the authors of the original datasets for making their data publicly available.

Contact

If you have questions or feedback, please contact:

• Jonah Kömen (koemen@tu-berlin.de)
• Edwin D. de Jong (edwin.dejong@aignostics.com)
• Julius Hense (j.hense@tu-berlin.de)

How to cite

If you find PathoROB useful, please cite our preprint:

@article{koemen2025pathorob,
    title={Towards Robust Foundation Models for Digital Pathology},
    author={K{\"o}men, Jonah and de Jong, Edwin D and Hense, Julius and Marienwald, Hannah and Dippel, Jonas and Naumann, Philip and Marcus, Eric and Ruff, Lukas and Alber, Maximilian and Teuwen, Jonas and others},
    journal={arXiv preprint arXiv:2507.17845},
    year={2025}
}

Please also cite the source publications of all PathoROB datasets that you use:

• Camelyon (Source: CAMELYON16 and CAMELYON17, License: CC0 1.0)

@article{bejnordi2017camelyon16,
    title={Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer},
    author={Ehteshami Bejnordi, Babak and Veta, Mitko and Johannes van Diest, Paul and van Ginneken, Bram and Karssemeijer, Nico and Litjens, Geert and van der Laak, Jeroen A. W. M. and and the CAMELYON16 Consortium},
    journal={JAMA},
    year={2017},
    volume={318},
    number={22},
    pages={2199-2210},
    doi={10.1001/jama.2017.14585}
}

@article{bandi19camelyon17,
    title={From Detection of Individual Metastases to Classification of Lymph Node Status at the Patient Level: The CAMELYON17 Challenge},
    author={Bándi, Péter and Geessink, Oscar and Manson, Quirine and Van Dijk, Marcory and Balkenhol, Maschenka and Hermsen, Meyke and Ehteshami Bejnordi, Babak and Lee, Byungjae and Paeng, Kyunghyun and Zhong, Aoxiao and Li, Quanzheng and Zanjani, Farhad Ghazvinian and Zinger, Svitlana and Fukuta, Keisuke and Komura, Daisuke and Ovtcharov, Vlado and Cheng, Shenghua and Zeng, Shaoqun and Thagaard, Jeppe and Dahl, Anders B. and Lin, Huangjing and Chen, Hao and Jacobsson, Ludwig and Hedlund, Martin and Çetin, Melih and Halıcı, Eren and Jackson, Hunter and Chen, Richard and Both, Fabian and Franke, Jörg and Küsters-Vandevelde, Heidi and Vreuls, Willem and Bult, Peter and van Ginneken, Bram and van der Laak, Jeroen and Litjens, Geert},
    journal={IEEE Transactions on Medical Imaging}, 
    year={2019},
    volume={38},
    number={2},
    pages={550-560},
    doi={10.1109/TMI.2018.2867350}
}

• TCGA (Source: TCGA-UT, License: CC-BY-NC-SA 4.0)

@article{komura22tcga-ut,
    title={Universal encoding of pan-cancer histology by deep texture representations},
    author={Daisuke Komura and Akihiro Kawabe and Keisuke Fukuta and Kyohei Sano and Toshikazu Umezaki and Hirotomo Koda and Ryohei Suzuki and Ken Tominaga and Mieko Ochi and Hiroki Konishi and Fumiya Masakado and Noriyuki Saito and Yasuyoshi Sato and Takumi Onoyama and Shu Nishida and Genta Furuya and Hiroto Katoh and Hiroharu Yamashita and Kazuhiro Kakimi and Yasuyuki Seto and Tetsuo Ushiku and Masashi Fukayama and Shumpei Ishikawa},
    journal={Cell Reports},
    year={2022},
    volume={38},
    number={9},
    pages={110424},
    doi={10.1016/j.celrep.2022.110424}
}

• Tolkach ESCA (Source: https://zenodo.org/records/7548828, License: CC-BY-SA 4.0)

@article{tolkach2023esca,
    title={Artificial intelligence for tumour tissue detection and histological regression grading in oesophageal adenocarcinomas: a retrospective algorithm development and validation study},
    author={Tolkach, Yuri and Wolgast, Lisa Marie and Damanakis, Alexander and Pryalukhin, Alexey and Schallenberg, Simon and Hulla, Wolfgang and Eich, Marie-Lisa and Schroeder, Wolfgang and Mukhopadhyay, Anirban and Fuchs, Moritz and others},
    journal={The Lancet Digital Health},
    year={2023},
    volume={5},
    number={5},
    pages={e265-e275},
    publisher={Elsevier}
}