# UCSF-PDGM: preoperative diffuse glioma MRI with tumor segmentations (BraTS-format structural subset)

## Facts

- Provider: Evan Calabrese et al., University of California San Francisco (converted by MedOtter)
- License: CC-BY-4.0 (https://www.cancerimagingarchive.net/collection/ucsf-pdgm/)
- Size: 2,480 files
- Format: NIfTI, CSV (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/UCSF-PDGM
- Access: public, readable without an account
- DOI: https://doi.org/10.7937/tcia.bdgf-8v37
- Category: Health & Biomedical
- Source: https://doi.org/10.1148/ryai.220058
- Page: https://scigantic.com/archive/07ecad19-853c-5461-94e6-62242617659a
- Keywords: Health & Biomedical, MRI, histopathology, image segmentation, NIfTI, CSV, Hugging Face

## About

The University of California San Francisco Preoperative Diffuse Glioma MRI (UCSF-PDGM) dataset, published on The Cancer Imaging Archive, is a cohort of preoperative 3 T brain MRI from patients with histopathologically confirmed diffuse gliomas, with molecular and clinical annotation. This MedOtter repository is the BraTS-format structural subset: for each of 495 patients a folder with T1, post-contrast T1c, T2 and FLAIR volumes co-registered to the 240 x 240 x 155 BraTS grid (1 mm isotropic, skull-stripped) and a tumor segmentation with BraTS labels (necrosis or non-enhancing core, edema, enhancing tumor), produced by an ensemble of BraTS algorithms and corrected by neuroradiologists. The diffusion, perfusion, SWI and HARDI series of the full TCIA release are not included. UCSF-PDGM-metadata_v5.csv gives sex, age, WHO CNS grade, WHO 2021 diagnosis, IDH, MGMT, 1p/19q, survival, extent of resection and BraTS21 IDs; about 60 percent of cases overlap BraTS 2021, so exclude them when evaluating alongside BraTS. As measured 2026-11-14 the repository has 495 patient folders of NIfTI (.nii.gz) files (about 6.2 GB in all); read them with nibabel or SimpleITK and the metadata with pandas. TCIA releases the collection under CC BY 4.0; cite Calabrese et al., Radiology: Artificial Intelligence 2022.

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/MedOtter/UCSF-PDGM/UCSF-PDGM-metadata_glossary.csv")
df.head()
```

## Data structure

File types in sampled files: NIfTI, CSV

Example file: `UCSF-PDGM-metadata_glossary.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/07ecad19-853c-5461-94e6-62242617659a
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/07ecad19-853c-5461-94e6-62242617659a
- Over MCP: `get_schema_card` with id `07ecad19-853c-5461-94e6-62242617659a` at https://api.scigantic.com/mcp

## Related topics

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