Skip to content
Open Imaging Index
MRI Brain

EGD

The Erasmus Glioma Database

Brain MRI before surgery of adults with glioma, paired with IDH, 1p/19q and grade labels that give the WHO 2016 subtype and with whole-tumor masks, for relating tumor appearance to genetics and for segmentation. 774 patients from routine care at Erasmus MC.

Go to dataset Edit Report

Overview

The Erasmus Glioma Database holds preoperative structural brain MRI of 774 adults with diffuse glioma who were treated at Erasmus MC in Rotterdam between 2008 and 2018. Each patient comes with four registered and defaced scans, a whole-tumor mask and, where known, the IDH mutation status, the 1p/19q co-deletion status and the tumor grade. The data were released to study how tumor appearance on MRI relates to genetic and histological features, and to build segmentation methods. In 2023 the Radboudumc organizers of the AUC23 classification challenge repackaged EGD cases as 4D MetaImage files labelled with WHO grade 2, 3 or 4 (Zenodo record 8054721, files restricted).

Composition

There are 281 women, 492 men and one patient of unknown sex, aged 19 to 86. IDH status is known for 467 patients, 1p/19q status for 259 and grade for 716. Combined, these give the full WHO 2016 subtype for 415 patients. Age, sex, scanner details and labels come as spreadsheets and as a JSON file per patient.

Acquisition

Scans come from routine clinical care. Every patient has a pre-contrast T1-weighted, a post-contrast T1-weighted, a T2-weighted and a T2-weighted FLAIR scan. Scanners came from Siemens, Philips, GE and Toshiba at 0.5 T to 3 T, with 1.5 T the most common. All scans were converted to NIfTI, affinely registered to the MNI152 atlas with Elastix and defaced with an atlas mask. A brain mask made with HD-BET is included for skull stripping.

Annotations

374 patients have a manual whole-tumor mask drawn by one of four observers on the T2-weighted or FLAIR scan. The other 400 have a mask produced by a convolutional neural network trained on manual masks. A spreadsheet records which observer or the network made each mask.

Known limitations

Molecular and grade labels are incomplete, so the full subtype is missing for almost half the cohort. Protocols and field strengths vary widely, and the field strength is unknown for four patients. The authors flag some scans with poor quality or failed registration as unacceptable on the XNAT server. Masks cover the whole tumor only, without subregions. All patients come from one center.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 774 subjects.

Sex

  • Female 281 36%
  • Male 492 64%
  • Unknown 1 0%

Abstract in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Condition

Groups can overlap

  • Glioma 774 100%

Abstract in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Scanner vendor

  • Siemens Healthineers 347 45%
  • Philips 254 33%
  • GE HealthCare 172 22%
  • Canon Medical 1 <1%

Methods: Imaging in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Country

  • Netherlands 774 100%

Specifications Table in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Age

range 19 to 86

No age bins reported.

Abstract in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Contrast / sequence

Groups can overlap

  • T1-weighted 774 100%
  • T1-weighted, contrast enhanced 774 100%
  • T2-weighted 774 100%
  • FLAIR 774 100%

Data Description: Structural MRI scans in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Field strength

  • 1.5 T 571 74%
  • 1 T 110 14%
  • 3 T 83 11%
  • 0.5 T 6 <1%

Methods: Imaging in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

Contrast combinations

How many subjects have exactly each set of contrasts.

T1wT1w_ceT2wFLAIRSubjects with exactly this set
774

Data Description: Structural MRI scans in van der Voort et al. 2021, Data in Brief (EGD data descriptor)

License and access

Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer

Access
Signed agreement

Sign a data use agreement, often reviewed by the provider

Access requests go by email to the Health-RI service desk (servicedesk@health-ri.nl), not through the XNAT portal. Applicants sign the data usage agreement and then receive an XNAT account. A Docker downloader is published at https://doi.org/10.5281/zenodo.4761088.

Access page

Erasmus Glioma Database Data License

A modified CC BY-NC-SA 4.0 license that forbids passing the data on. Non-commercial use only, adaptations must keep the same license elements, and access requires signing the agreement with Health-RI.

Original license text Version read: Supplementary file of van der Voort et al., Data in Brief 2021, read 2026-10-11 Checked 2026-10-11

What you can do

  • No
  • Not stated
  • Not stated
  • Conditional
  • Not stated

What you can share

  • No
  • Conditional
  • Not stated

What you must do

  • Yes
  • Yes
  • Yes
  • No
  • Manuscript review No
  • Release code No
  • Return results No
  • Delete after use No

Limits

  • No
  • Location limits No

Citation

van der Voort SR, Incekara F, Wijnenga MMJ, Kapsas G, Gahrmann R, Schouten JW, Dubbink HJ, Vincent AJPE, van den Bent MJ, French PJ, Klein S, Smits M. The Erasmus Glioma Database (EGD): Structural MRI scans, WHO 2016 subtypes, and segmentations of 774 patients with glioma. Data in Brief 37, 107191 (2021). https://doi.org/10.1016/j.dib.2021.107191

Sources

Every number on this page comes from one of these documents. Each chart names the table or page it is taken from. The raw numbers are in stats.csv.