The UNI2-h-DSS datasets used in the Path-PKT study are released under the CC-BY-NC-ND 4.0 license and may only be used for non-commercial, academic research purposes with proper attribution. Any commercial use, sale, or other monetization of the datasets and their derivatives is prohibited and requires prior approval. By downloading the datasets, you attest that all information (affiliation, research use) is correct and up-to-date. Downloading the datasets requires prior registration on Hugging Face and agreeing to the terms of use. By downloading the datasets, you agree not to distribute, publish or reproduce a copy of the datasets. If another user within your organization wishes to use the datasets, they must register as an individual user and agree to comply with the terms of use. If you are a commercial entity, please contact the author ([email protected]).

Dataset Card for UNI2-h-DSS

Name: UNI2-h-DSS
Creator: Pei Liu
License: https://choosealicense.com/licenses/cc-by-nc-nd-4.0/

This dataset card provides the details of the UNI2-h-DSS dataset used in Path-PKT (Liu et al., 2025). DSS refers to Disease-Specific Survival, which is a typical event of interest in survival analysis.

Requesting Access

As mentioned above, you must agree to the terms of use. An institutional email (e.g., EDU email address) is preferred.

Dataset Description

UNI2-h-DSS is a dataset used in the Path-PKT study. It is curated for the study of WSI-based cancer prognosis. It has

13 datasets for cancer-specific training, with 7,268 patients and 8,818 WSIs in total;
13 datasets (each corresponding to one rare tumor disease) for algorithm validation, with 1,922 patients and 2,370 WSIs.

Content

data-splits-for-cancer-specific-training: Data split files (5-fold cross-validation) for 13 cancer datasets used in cancer-specific training.
follow-up-labels-of-rare-tumor-datasets: DSS follow-up labels of all patients in 13 rare tumor datasets.
Distributed 'CSV' files: they contain the information of patient_id, slide_id, censorship status, and DSS follow-up time associated with all cancer patients.

Original WSIs can be downloaded at TCGA GDC portal; The image features extracted by UNI2-h (Chen et al., 2024) can be accessed here.

Information Summary

Original WSIs: TCGA GDC portal
WSI features: Please download them here
Repository: https://github.com/liupei101/Path-PKT
Paper: arXiv Preprint (https://arxiv.org/abs/2508.13482)
License: CC-BY-NC-ND-4.0

Usage

Following authentication, the datasets can be loaded manually at this website or in the following ways:

Install huggingface-cli and log into your account

pip install -U "huggingface_hub[cli]"

huggingface-cli login

Download datasets

local_dir=/path/to/local_dir
huggingface-cli download yuukilp/UNI2-h-DSS --repo-type dataset --local-dir $local_dir

Contact

For any additional questions or comments, please contact Pei Liu ([email protected]).

How to Cite

Anyone who uses this dataset should consider cite the following works:

Path-PKY study (curating 26 cancer-specific datasets with DSS follow-up labels for cancer prognosis research):

@misc{liu2025pathpkt,
      title={Towards Understanding and Harnessing the Transferability of Prognostic Knowledge in Computational Pathology}, 
      author={Pei Liu and Luping Ji and Jiaxiang Gou and Xiangxiang Zeng},
      year={2025},
      eprint={2508.13482},
      archivePrefix={arXiv},
      url={https://arxiv.org/abs/2508.13482}, 
}

Original TCGA datasets (providing pan-cancer cohorts):

@article{weinstein2013cancer,
  title={The cancer genome atlas pan-cancer analysis project},
  author={Weinstein, John N and Collisson, Eric A and Mills, Gordon B and Shaw, Kenna R and Ozenberger, Brad A and Ellrott, Kyle and Shmulevich, Ilya and Sander, Chris and Stuart, Joshua M},
  journal={Nature genetics},
  volume={45},
  number={10},
  pages={1113--1120},
  year={2013},
  publisher={Nature Publishing Group}
}

Pretrained UNI2-h models (for image feature extraction):

@article{chen2024uni,
  title={Towards a General-Purpose Foundation Model for Computational Pathology},
  author={Chen, Richard J and Ding, Tong and Lu, Ming Y and Williamson, Drew FK and Jaume, Guillaume and Chen, Bowen and Zhang, Andrew and Shao, Daniel and Song, Andrew H and Shaban, Muhammad and others},
  journal={Nature Medicine},
  publisher={Nature Publishing Group},
  year={2024}
}

Datasets:

yuukilp
/

UNI2-h-DSS

You need to agree to share your contact information to access this dataset