---
license: other
pretty_name: MDS-Bench Raw Data
task_categories:
  - image-segmentation
  - image-classification
  - object-detection
tags:
  - medical
  - medical-imaging
  - benchmark
viewer: false
---

# MDS-Bench Raw Data

Raw input data for [MDS-Bench](https://github.com/chenxin-sdu/MDS-Bench), a benchmark that tests whether
AI agents can organize heterogeneous medical imaging datasets into a unified image + JSON format.

It contains **100 public medical imaging datasets** (CT, MRI, microscopy, X-ray, ultrasound, endoscopy, ...;
DICOM, NIfTI, TIFF, H5, MAT, ...), about 2 TB in total.

## Layout

```
origin_dataset/<dataset_name>/data/...   # 8 datasets
second_dataset/<dataset_name>/data/...   # 34 datasets
third_dataset/<dataset_name>/data/...    # 22 datasets
forth_dataset/<dataset_name>/data/...    # 22 datasets
fifth_dataset/<dataset_name>/data/...    # 14 datasets
```

Numeric `dataset_name`s are Zenodo record IDs.

## Download

```bash
hf download sdu-chenxin/MDS-Bench-data --repo-type dataset \
  --include "fifth_dataset/LCTSC/data/*" --local-dir ./MDS-Bench-data
```

## Notes

- `second_dataset/fomo-mri`: because of the Hub limit of 1,000,000 files per repository, part of
  `fomo-60k/` is packed into `fomo-60k-shards/shard_000.tar` ... `shard_038.tar`. Extract them into the same
  `data/` folder (`tar -xf shard_XXX.tar -C data/`) to restore the full dataset; `INDEX.tsv` lists which shard
  each file is in.
- `third_dataset/brats2020-training-data` and `fifth_dataset/rsna-pneumonia-detection-challenge` have some
  files moved into subfolders because of the Hub limit of 10,000 files per folder.

## License

Each dataset keeps its original license and terms of use. Please check the original source and cite the
original authors when using a dataset.