# TOM500: orbital MRI of 500 thyroid eye disease patients with nine-structure segmentations

## Facts

- Provider: Shanghai Ninth People's Hospital TOM500 team; mirrored by MedOtter
- License: CC0-1.0 (https://api.figshare.com/v2/articles/27133389)
- Size: 1,004 files
- Format: NIfTI, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/TOM500
- Access: public, readable without an account
- DOI: https://doi.org/10.6084/m9.figshare.27133389
- Category: Health & Biomedical
- Source: https://doi.org/10.1038/s41597-025-04427-9
- Page: https://scigantic.com/archive/b994f551-8af8-5564-a99e-a9065df5b3b3
- Keywords: Health & Biomedical, MRI, image segmentation, NIfTI, Parquet, Hugging Face

## About

TOM500 is a multi-organ annotated orbital MRI dataset for thyroid eye disease (TED, Graves' orbitopathy). It contains coronal T2-weighted DRIVE MRI of 500 patients with clinically confirmed TED, acquired at Shanghai Ninth People's Hospital between June 2016 and January 2021 on a 3.0 T Philips Ingenia with a 32-channel head coil; volumes are 512x512x20 with 0.3125 mm in-plane resolution and 3.5 mm slices. Masks label nine periocular structures: optic nerve, orbital fat, lacrimal gland, eyeball, the superior rectus and levator palpebrae superioris complex, inferior, medial and lateral recti, and superior oblique. Each case was annotated by three junior radiologists, fused with STAPLE and reviewed by a senior expert. Split 400 training and 100 validation; NIfTI files under TOM500/, about 2.7 GB (measured 2026-11-17). Use for extraocular muscle segmentation and TED severity assessment. The source figshare record also has a clinical CSV (sex, age, disease duration, smoking). Mirrored on Hugging Face by MedOtter from the authors' release, with a card documenting provenance and any normalization. The figshare record 27133389 is CC0; cite the Scientific Data article (doi 10.1038/s41597-025-04427-9).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("MedOtter/TOM500")
print(ds)
```

## Data structure

File types in sampled files: NIfTI, Parquet

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

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/b994f551-8af8-5564-a99e-a9065df5b3b3
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/b994f551-8af8-5564-a99e-a9065df5b3b3
- Over MCP: `get_schema_card` with id `b994f551-8af8-5564-a99e-a9065df5b3b3` at https://api.scigantic.com/mcp

## Related topics

- [MRI datasets](https://scigantic.com/topics/mri)
- [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)
- [NIfTI datasets](https://scigantic.com/topics/nifti)
