# WAW-TACE: multiphase CT of 233 hepatocellular carcinoma (HCC) patients treated with trans-arterial chemoembolization (TACE)

## Facts

- Provider: Krzysztof Bartnik, Tomasz Bartczak, Mateusz Krzyzinski et al. (Medical University of Warsaw); mirrored by MedOtter
- License: CC-BY-4.0 (https://zenodo.org/records/12741586)
- Size: 2,092 files
- Format: Parquet, NIfTI (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/WAW-TACE
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.12741586
- Category: Health & Biomedical
- Source: https://zenodo.org/records/12741586
- Page: https://scigantic.com/archive/ee0a9d5f-10f9-549e-bc81-9eeb6c703a97
- Keywords: Health & Biomedical, CT, image segmentation, Parquet, NIfTI, Hugging Face

## About

WAW-TACE (Warsaw TACE) is a multiphase abdominal CT dataset from the Medical University of Warsaw of 233 treatment-naive hepatocellular carcinoma (HCC, primary liver cancer) patients who received trans-arterial chemoembolization (TACE) as monotherapy. It has 854 CT volumes across four contrast phases (200 precontrast, 230 late arterial, 231 portal venous, 193 delayed), 378 hand-drawn expert-validated liver tumor masks, TotalSegmentator-generated masks of 104 abdominal structures for every volume (not reviewed by hand), radiomics features and clinical outcome data. Folders are images/, tumor_masks/, organ_masks/ and metadata/ (per-patient phase availability and clinical tables), with phase encoded as an integer suffix in file names; NIfTI throughout, about 48 GB (measured 2026-11-21). Read with nibabel or SimpleITK. Use for liver tumor segmentation across contrast phases, TACE response and survival prediction, and radiomics studies. Mirrored by MedOtter from the authors' Zenodo record, released under CC BY 4.0 with DOI 10.5281/zenodo.12741586; cite Bartnik et al. and the Zenodo record.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("MedOtter/WAW-TACE")
print(ds)
```

## Data structure

File types in sampled files: Parquet, NIfTI

Example file: `data/train-00000-of-00001.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/ee0a9d5f-10f9-549e-bc81-9eeb6c703a97
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/ee0a9d5f-10f9-549e-bc81-9eeb6c703a97
- Over MCP: `get_schema_card` with id `ee0a9d5f-10f9-549e-bc81-9eeb6c703a97` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [Health & Biomedical datasets](https://scigantic.com/topics/health-and-biomedical)
- [CT datasets](https://scigantic.com/topics/ct)
- [NIfTI datasets](https://scigantic.com/topics/nifti)
