Skip to content
Open Imaging Index
MRI Brain

EPISURG

EPISURG: a dataset of postoperative magnetic resonance images (MRI) for quantitative analysis of resection neurosurgery for refractory epilepsy

Postoperative brain MRI of people with drug-resistant focal epilepsy after resective surgery, released to train and evaluate automatic segmentation of the resection cavity. Includes preoperative scans for most patients and manual cavity masks from three raters on a subset.

Go to dataset Edit Report

Overview

EPISURG holds T1-weighted brain MRI of patients with refractory focal epilepsy who had part of the brain removed in epilepsy surgery at the National Hospital for Neurology and Neurosurgery in London between 1990 and 2018. The group of Sébastien Ourselin and John Duncan at University College London released it to develop and test methods that segment the resection cavity on postoperative scans, so that the removed tissue can be mapped back to a preoperative image and its parcellation.

Composition

The release contains one postoperative T1-weighted image for each of 430 patients. 269 of them also have the matching preoperative T1-weighted image, for 699 images in total. The 2020 MICCAI paper reports 431 patients, and the 2021 IJCARS paper reports 268 preoperative images; the counts here follow the repository record. No age, sex or surgery-type breakdown is published.

Acquisition

The images come from routine clinical care at a single hospital and were collected retrospectively as anonymised data, so no individual consent was sought. The sources do not list scanners, field strengths or sequence parameters. Files are NIfTI and every image was defaced with a fixed face mask defined in MNI space.

Annotations

Three raters outlined the resection cavity on overlapping subsets of the postoperative images: a neuroimaging researcher on 133 patients, a clinical scientist on 34 and a neurologist on 33. In the MICCAI paper the 133 were split into ten subsets with a similar mix of resection types (for example temporal or frontal); the second and third raters each covered two of those subsets, which gives an estimate of inter-rater agreement.

Known limitations

  • Only T1-weighted images are included; other clinical sequences are not part of the release.
  • Scanner and protocol details are not documented, and the scans span almost three decades.
  • Roughly a third of patients have no preoperative image.
  • Most postoperative images (297 of 430) have no manual cavity mask.
  • The data use terms forbid passing the data on in any form, which goes further than the CC BY-NC-SA 4.0 license on the record.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 430 subjects.

Contrast / sequence

Groups can overlap

  • T1-weighted 430 100%

From EPISURG record on the UCL Research Data Repository (v1)

Country

  • United Kingdom 430 100%

From EPISURG record on the UCL Research Data Repository (v1)

Condition

Groups can overlap

  • Epilepsy 430 100%

From EPISURG record on the UCL Research Data Repository (v1)

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
Open download

Download without an account

A single EPISURG.zip downloads from the UCL Research Data Repository without an account. The record carries a CC BY-NC-SA 4.0 license, and its description adds data use terms that every user must follow.

Access page

Different parts of the data carry different licenses. The summary on the right shows the most restrictive answer per rule.

For The whole UCL Research Data Repository record (license stated in the record metadata)

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International

Non-commercial use only, with credit. Anything you share that is built on the data must use the same license.

Original license text Version read: 4.0 Checked 2026-10-07

What you can do

  • No
  • Not stated
  • Conditional
  • Yes
  • Yes

What you can share

  • Conditional
  • Conditional
  • Conditional

What you must do

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

Limits

  • No
  • Location limits No
For The whole record (terms stated in the record description)

EPISURG data use agreement (UCL Research Data Repository record description)

Four terms printed on the download page, accepted by using the data without any signature. Users must not try to identify or contact patients, must cite the two listed publications, must not pass the data on to anyone, and must make their team follow the same terms.

Original license text Version read: Four numbered terms in the description of record 9996158 v1, read 2026-10-11 Checked 2026-10-11

What you can do

  • Not stated
  • Not stated
  • Conditional
  • Conditional
  • Conditional

What you can share

  • No
  • Not stated
  • Share trained models Not stated

What you must do

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

Limits

  • Yes
  • No

Commercial license: Not stated

Citation

Pérez-García F, Rodionov R, Alim-Marvasti A, Sparks R, Duncan JS, Ourselin S. Simulation of Brain Resection for Cavity Segmentation Using Self-supervised and Semi-supervised Learning. MICCAI 2020, LNCS 12263, Springer (2020). doi:10.1007/978-3-030-59716-0_12. And: Pérez-García F, Rodionov R, Alim-Marvasti A, Sparks R, Duncan JS, Ourselin S. EPISURG: MRI dataset for quantitative analysis of resective neurosurgery for refractory epilepsy. University College London (2020). doi:10.5522/04/9996158.v1

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.