# Arctique: synthetic H&E-like colon tissue images with nucleus instance, nucleus class and cytoplasm masks

## Facts

- Provider: Jannik Franzen, Claudia Winklmayr, Vanessa Emanuela Guarino et al.; curated by SPORA
- License: MIT (https://zenodo.org/records/14016860)
- Size: 8 files
- Format: JSON, PNG, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/csmp-hub/arctique
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.14016860
- Category: Health & Biomedical
- Source: https://arxiv.org/abs/2411.07097
- Page: https://scigantic.com/archive/fa036556-a28d-5ee8-a323-96ba7fd23c63
- Keywords: Health & Biomedical, histopathology, image segmentation, JSON, PNG, Parquet, Hugging Face

## About

Arctique (ARtificial Colon Tissue Images for Qualitative Uncertainty Evaluation) is a synthetic dataset of 512x512 images that resemble H&E-stained colon tissue, generated so that ground truth masks are known exactly. Each image comes with a nucleus instance mask, a nucleus class mask and a cytoplasm class mask; classes are epithelial cell, plasma cell, lymphocyte, eosinophil and fibroblast, with cytoplasm labeled for plasma cells and eosinophils. This package holds the base images of the Zenodo v3 experimental archive, keeping the original 1,450 training and 50 test images with original file bytes (image as RGBA PNG, masks as TIFF); the source archive also offers noisy variants, depth maps and 3D masks. About 780 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. Intended for benchmarking segmentation and uncertainty quantification under controlled label noise and distribution shift. Curated and packaged on Hugging Face by the SPORA project (csmp-hub), which keeps the original license and citation files. The Zenodo release 14016860 is under the MIT License; cite Franzen, Winklmayr, Guarino et al., and their NeurIPS paper (arXiv 2411.07097).

## Load the data

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

## Data structure

File types in sampled files: JSON, PNG, Parquet

Example file: `test.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/fa036556-a28d-5ee8-a323-96ba7fd23c63
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/fa036556-a28d-5ee8-a323-96ba7fd23c63
- Over MCP: `get_schema_card` with id `fa036556-a28d-5ee8-a323-96ba7fd23c63` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [JSON datasets](https://scigantic.com/topics/json)
- [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)
