# UnumLocalia: same-section H&E, Xenium 5,001-gene spatial transcriptomics and COMET protein imaging of liver and tonsil

## Facts

- Provider: Duchini et al. and Felix Marsh-Wakefield; curated by SPORA
- License: CC-BY-4.0 (https://zenodo.org/records/21713660)
- Size: 15 files
- Format: CSV, Parquet, PNG (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/csmp-hub/unumlocalia
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.21713660
- Category: Health & Biomedical
- Source: https://doi.org/10.64898/2026.08.17.742355
- Page: https://scigantic.com/archive/58ab6fa0-2b94-5908-a577-ddd391f88883
- Keywords: Health & Biomedical, histopathology, image segmentation, spatial transcriptomics, CSV, Parquet, PNG, Hugging Face

## About

A matched multimodal resource from four human tissue cores (hepatocellular carcinoma, matched non-tumor liver, tonsil and hepatocellular adenoma) in which H&E, Xenium in situ spatial transcriptomics with a 5,001-target RNA panel, and COMET sequential immunofluorescence with 15 protein markers plus DAPI were all measured on the same tissue section. This package crops the H&E into 513 tiles of 512x512 pixels at about 0.203 micrometers per pixel with unchanged pixel values, and converts the supplied Xenium cell and nucleus boundaries into instance masks (original numeric IDs preserved, nucleus-to-cell pairs listed). The full source has 46,076 cells and 45,891 nuclei; objects.parquet holds every boundary and link, and measurements/ carries four further measurement files from the source. About 1.6 GB (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. Use for predicting gene or protein expression from H&E, multimodal registration and cell segmentation benchmarks. Curated and packaged on Hugging Face by the SPORA project (csmp-hub), which keeps the original license and citation files. The original Zenodo release 21713660 is CC BY 4.0; credit Duchini and colleagues and Felix Marsh-Wakefield and cite the study (doi 10.64898/2026.08.17.742355).

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/csmp-hub/unumlocalia/class-distribution.csv")
df.head()
```

## Data structure

File types in sampled files: CSV, Parquet, PNG

Example file: `class-distribution.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/58ab6fa0-2b94-5908-a577-ddd391f88883
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/58ab6fa0-2b94-5908-a577-ddd391f88883
- Over MCP: `get_schema_card` with id `58ab6fa0-2b94-5908-a577-ddd391f88883` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [CSV datasets](https://scigantic.com/topics/csv)
- [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)
- [Spatial transcriptomics datasets](https://scigantic.com/topics/spatial-transcriptomics)
