README.md

---
license: cc-by-nc-nd-4.0
extra_gated_prompt: >-
  The TCGA-UniformTumor-8K dataset is 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 dataset and their derivatives are prohibited and requires prior approval.
  Please note that the primary email used to sign up for your Hugging Face account must match your institutional email to receive approval. By downloading the features, you attest that all information (affiliation, research use) is correct and up-to-date.
  Downloading the dataset requires prior registration on Hugging Face and agreeing to the terms of use.
  By downloading the dataset, you agree not to distribute, publish or reproduce a copy of the data.
  If another user within your organization wishes to use the data, 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 corresponding author.
extra_gated_fields:
  Full name (first and last): text
  Current affiliation (no abbreviations): text
  Type of Affiliation:
    type: select
    options: 
      - Academia
      - Industry
      - label: Other
        value: other
  Current and official institutional email (**this must match your primary email in your Hugging Face account, @gmail/@hotmail/@qq email domains will be denied**): text
  Please explain your intended research use: text
  I agree to all terms outlined above: checkbox
  I agree to use this model for non-commercial, academic purposes only: checkbox
  I agree not to distribute the model, if another user within your organization wishes to use the released pretrained features, they must register as an individual user: checkbox
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
dataset_info:
  features:
  - name: PATIENT
    dtype: string
  - name: FILENAME
    dtype: string
  - name: image
    dtype: image
  - name: x
    dtype: int64
  - name: y
    dtype: int64
  - name: cancer
    dtype: string
  - name: source_site
    dtype: string
  - name: source
    dtype: string
  splits:
  - name: train
    num_bytes: 186375931310.701
    num_examples: 13853
  - name: val
    num_bytes: 49282909947.13
    num_examples: 3434
  - name: test
    num_bytes: 118463491347.128
    num_examples: 8208
  download_size: 376933091793
  dataset_size: 354122332604.959
---

# Dataset Card for TCGA-UniformTumor-8K

#### What is TCGA-UniformTumor-8K?

TCGA-UniformTumor-8K dataset is a region-level pan-cancer subtyping resource comprising 25,495 ROIs of 8,192 × 8,192 pixels. These regions were extracted from 9,662 H&E-stained FFPE diagnostic histopathology WSIs sourced from TCGA. The tumor regions were manually annotated by two expert pathologists, with slide exclusion due to poor staining, poor focus, lacking cancerous regions and incorrect cancer types. Approximately three representative tumor regions per WSI were annotated with pixel-level contours. For each contour, we center-cropped an image region of 8,192 × 8,192 pixels to encompass both the dense tumor and its surrounding tissue context.

<img src="tcga_ut_examples.jpg" alt="Description" style="width: 75%;"/>

## Instructions for Downloading TCGA-UniformTumor-8K

### Install huggingface-hub and datasets, and log into your account

```bash
pip install huggingface-hub
pip install datasets
```

```bash
from huggingface_hub import login
login(token="YOUR HUGGINGFACE TOKEN")
```

### Download TCGA-UniformTumor-8K

```bash
from datasets import load_dataset
ds = load_dataset("MahmoodLab/TCGA-UniformTumor-8K")     # returns a DatasetDict
# example:
img, row = ds["train"][0]["image"], ds["train"][0]       # PIL image + image metadata
```

### How to cite:

```
@article{ding2025multimodal,
  title={A multimodal whole-slide foundation model for pathology},
  author={Ding, Tong and Wagner, Sophia J and Song, Andrew H and Chen, Richard J and Lu, Ming Y and Zhang, Andrew and Vaidya, Anurag J and Jaume, Guillaume and Shaban, Muhammad and Kim, Ahrong and others},
  journal={Nature Medicine},
  pages={1--13},
  year={2025},
  publisher={Nature Publishing Group US New York}
}
```

### Contact:
- <b>Tong Ding</b> (`tong_ding@g.harvard.edu`)
- <b>Sophia J. Wagner</b> (`sjwagner@bwh.harvard.edu`)
- <b>Andrew H. Song</b> (`asong@bwh.harvard.edu`)
- <b>Richard J. Chen</b> (`richardchen@g.harvard.edu`)
- <b>Faisal Mahmood</b> (`faisalmahmood@bwh.harvard.edu`)