Pre-operative glioma segmentation ยท multimodal MRI
๐ Publications and records
| Quick facts | |
|---|---|
| Input | T1, T1ce, T2, FLAIR |
| Task | Glioma subregion segmentation and related challenge tasks |
| Labels | NCR/NET, edema, enhancing tumor; derived regions WT, TC, ET |
| Institutional origin | Challenge coordination was led through CBICA at the Perelman School of Medicine, University of Pennsylvania (UPenn). The official challenge describes the pre-operative MRI cohort as multi-institutional. Official challenge page. |
| Official source | Challenge page ยท Official data notes |
๐งพ Dataset overview
BraTS 2019 uses multi-institutional, pre-operative MRI and expert-reviewed tumor annotations. The official data notes describe the standardized, co-registered and skull-stripped images and explain the annotation protocol.
The challenge training release contains 335 labeled cases. The official challenge also used validation and testing phases; the challenge page says test data and evaluation were released only to submitted papers during a short post-submission window.
๐๏ธ Official split and label availability
| Cohort | Official status |
|---|---|
| Training | 335 cases with reference segmentations |
| Validation | Challenge evaluation cohort; reference labels were not part of the ordinary public training release |
| Test | Challenge evaluation cohort; the 2019 page documents a restricted, time-limited release to submitted papers rather than unrestricted public test labels |
The official challenge page does not make the old challenge split equivalent to a modern, permanent public test-label release. Check the original organizer source for current access.
๐ท๏ธ Labels and evaluation notes
The official 2019 data page specifies voxel labels 1 = NCR/NET, 2 = edema, and 4 = enhancing tumor. The common derived regions are WT = {1,2,4}, TC = {1,4}, and ET = {4}. Keep these definitions attached to reported metrics.
๐ Access and terms
The original challenge page remains the source for its release schedule and access terms. The legacy Synapse link may not provide the historical files today; do not substitute an unofficial mirror without checking provenance and terms.
๐ Related challenge
BraTS 2019, held with MICCAI 2019 as part of the BrainLes workshop, evaluated glioma MRI segmentation, overall-survival prediction, and segmentation uncertainty.
๐ Publications and citation
Challenge proceedings: BrainLes 2019, held in conjunction with MICCAI 2019 and published in 2020 as Springer Lecture Notes in Computer Science, vols. 11992 (Part I) and 11993 (Part II).
- Part I, LNCS 11992, DOI: 10.1007/978-3-030-46640-4 ยท Part II, LNCS 11993, DOI: 10.1007/978-3-030-46643-5
Dataset resource paper: Bakas et al. โAdvancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features.โ Scientific Data, vol. 4, article 170117, 2017. DOI: 10.1038/sdata.2017.117.
Foundational benchmark reference: Menze et al. โThe Multimodal Brain Tumor Image Segmentation Benchmark (BRATS).โ IEEE Transactions on Medical Imaging, vol. 34, no. 10, pp. 1993โ2024, 2015. DOI: 10.1109/TMI.2014.2377694.
The MICCAI 2019 challenge archive and split/label details are linked in the official source section above.
๐ Sources last checked: 27 October 2026 ยท Back to the dataset catalog