# CytoNuke: whole-cell and nucleus outlines in H&E images of head and neck squamous cell carcinoma

## Facts

- Provider: Johannes Raufeisen, Kunpeng Xie, Fabian Hörst et al.; curated by SPORA
- License: CC-BY-4.0 (https://zenodo.org/records/15310323)
- Size: 13 files
- Format: PNG, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/csmp-hub/cytonuke
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.15310323
- Category: Health & Biomedical
- Source: https://doi.org/10.1016/j.cmpb.2024.108215
- Page: https://scigantic.com/archive/d7b46e42-9216-5c30-a58b-50308a15b5b8
- Keywords: Health & Biomedical, histopathology, image segmentation, PNG, Parquet, Hugging Face

## About

CytoNuke addresses whole-cell (not just nucleus) segmentation in bright-field histology. It has 83 H&E images of head and neck squamous cell carcinoma (HNSCC) from 13 patients, each 256x256 pixels at a reported 0.5 micrometers per pixel, with hand-drawn outlines of nuclei and of the full cell boundary where it could be identified, so whole-cell masks cover only part of the labeled nuclei. The images come from the TCIA CPTAC-HNSCC collection with stain colors adjusted by the dataset creators. Patient-level splits: train.parquet (43 images, 9 patients, 2,344 nuclei, 1,547 whole cells), val.parquet (11 images, 2 patients) and test.parquet (29 images, 2 patients); these differ from the paper's evaluation groups. The original Zenodo archive and its README are included. About 30 MB (measured 2026-11-17). Images and masks are stored as PNG or TIFF bytes inside Parquet; read them with polars or pandas plus Pillow, or with Hugging Face datasets. Curated and packaged on Hugging Face by the SPORA project (csmp-hub), which keeps the original license and citation files. The creators' Zenodo record 15310323 is CC BY 4.0, and the CPTAC/TCIA source images are CC BY 3.0; keep both notices. Cite Raufeisen, Xie, Hörst et al., Computer Methods and Programs in Biomedicine 2024 (doi 10.1016/j.cmpb.2024.108215).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("csmp-hub/cytonuke")
print(ds)
```

## Data structure

File types in sampled files: PNG, Parquet

Example file: `test.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/d7b46e42-9216-5c30-a58b-50308a15b5b8
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/d7b46e42-9216-5c30-a58b-50308a15b5b8
- Over MCP: `get_schema_card` with id `d7b46e42-9216-5c30-a58b-50308a15b5b8` 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)
- [PNG datasets](https://scigantic.com/topics/png)
- [Histopathology datasets](https://scigantic.com/topics/histopathology)
