UPENN-GBM
University of Pennsylvania Glioblastoma Imaging, Genomics, and Radiomics
Multi-parametric brain MRI (T1, contrast-enhanced T1, T2, FLAIR, DTI, DSC) of 630 de novo glioblastoma patients from the University of Pennsylvania, with tumor sub-region segmentations, survival, IDH and MGMT data, radiomic features and a few H&E slides.
Overview
UPENN-GBM is a single-center collection of brain MRI from 630 adults with newly diagnosed (de novo) glioblastoma, treated at the University of Pennsylvania Health System between 2006 and 2018. It was assembled by the Penn imaging group behind CaPTk and BraTS and is hosted by The Cancer Imaging Archive. Besides the scans it ships clinical, molecular and radiomic tables, which makes it a common source for tumor segmentation, survival prediction and radiogenomic studies.
Composition
There are 671 scan sessions: 611 pre-operative baseline sessions and 60 sessions acquired before a second resection for suspected progression. For 19 of those 60 patients only the follow-up session is included. Every session has native T1, contrast-enhanced T1, T2 and FLAIR; 592 sessions also have DTI and 534 have DSC perfusion. Patients are 60% male and aged 18 to 89. The clinical table gives sex, age, overall survival (452 patients), extent of resection, IDH1 status (16 mutated, 499 wildtype, 96 not otherwise specified among baseline cases), MGMT promoter methylation (317 baseline cases) and, for follow-up cases, a pathology score of tumor progression versus treatment effect. Version 2 added 71 H&E whole-slide images from 34 patients, with a mapping to the radiology IDs.
Acquisition
Scans come from routine clinical exams, mostly on Siemens 3 T systems; a minority were acquired at 1.5 T and eight sessions on GE scanners. The original DICOM series are defaced. A NIfTI release is also provided in which all structural scans are rigidly registered to the SRI atlas, resampled to 1 mm isotropic voxels and skull-stripped, together with DTI maps (trace, axial and radial diffusivity, fractional anisotropy) and DSC maps (peak height, percentage signal recovery, an rCBV proxy).
Annotations
Enhancing tumor, necrotic core and peritumoral edema were segmented on every baseline session by fusing three BraTS-ranked networks with STAPLE. For 232 subjects the labels were reviewed and corrected by two experts. CaPTk radiomic features (145 per sub-region and sequence) are provided for the baseline sessions.
Known limitations
- Single institution and mostly one scanner vendor, so external validity is limited.
- Most segmentations are automatic and unreviewed; only 232 were manually corrected.
- MGMT and IDH status are missing for many patients and were not re-validated beyond the medical records.
- The NIfTI images are in atlas space and do not align with the DICOM images.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 630 subjects.
Sex
- Female 252 40%
- Male 378 60%
Age
mean 62.7± 12.5 · range 18 to 89Contrast combinations
How many subjects have exactly each set of contrasts.
| T1w | T1w_ce | T2w | FLAIR | dwi | dsc | Subjects with exactly this set |
|---|---|---|---|---|---|---|
491 | ||||||
70 | ||||||
63 | ||||||
6 |
Modality
subjects
- MRI 630 100%
- Pathology slides 34 5%
Contrast / sequence
subjects, values can overlap
- T1-weighted 630 100%
- T1-weighted, contrast enhanced 630 100%
- T2-weighted 630 100%
- FLAIR 630 100%
- Diffusion-weighted 554 88%
- DSC perfusion 497 79%
Condition
subjects, values can overlap
- Glioblastoma 630 100%
Scanner vendor
subjects
- Siemens Healthineers 622 99%
- GE HealthCare 8 1%
Field strength
studies
- 3 T 596 89%
- 1.5 T 74 11%
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
Download without an account
DICOM via the TCIA Data Retriever; pre-processed NIfTI images, segmentations and histopathology slides via IBM Aspera; clinical, acquisition and radiomic tables as direct CSV downloads.
Creative Commons Attribution 4.0 International
Use, share and adapt the data for any purpose, including commercial use, as long as you credit the creators.
Every item on the TCIA collection page is listed under CC BY 4.0. TCIA also asks users to follow its Data Usage Policy and to cite the dataset DOI.
What you can do
- Yes
- Yes
- Yes
- Yes
What you can share
- Yes
- Yes
- Yes
What you must do
- Yes
- Share alike No
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Citation
Bakas S, Sako C, Akbari H, et al. The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: advanced MRI, clinical, genomics, & radiomics. Scientific Data 9, 453 (2022). https://doi.org/10.1038/s41597-022-01560-7. Data: Bakas S, et al. (2021). Multi-parametric magnetic resonance imaging (mpMRI) scans for de novo Glioblastoma (GBM) patients from the University of Pennsylvania Health System (UPENN-GBM) (Version 2) [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/TCIA.709X-DN49
All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Subjects | total | 630 | bakas2022 Methods, Data |
| Subjects | condition=glioblastoma de novo glioblastoma | 630 | bakas2022 Abstract |
| Subjects | modality=MR | 630 | tcia-upenn-gbm Data Access (Version 2) |
| Subjects | modality=SM subjects with histopathology slides | 34 | tcia-upenn-gbm Data Access (Version 2) |
| Studies | total 611 pre-operative baseline and 60 follow-up scan sessions | 671 | bakas2022 Table 1 |
| Subjects | sex=female | 252 | bakas2022 Table 1 |
| Subjects | sex=male | 378 | bakas2022 Table 1 |
| Subjects | age=18-29 | 14 | bakas2022 Table 1 |
| Subjects | age=30-49 | 69 | bakas2022 Table 1 |
| Subjects | age=50-69 | 367 | bakas2022 Table 1 |
| Subjects | age=70+ | 180 | bakas2022 Table 1 |
| Minimum age | total age range 18-89 | 18 | bakas2022 Methods, Clinical data |
| Maximum age | total age range 18-89 | 89 | bakas2022 Methods, Clinical data |
| Mean age | total age at the first available scan of each of the 630 subjects | 62.7 | tcia-upenn-gbm-clinical ID, Age_at_scan_years |
| Median age | total age at the first available scan of each of the 630 subjects | 63.4 | tcia-upenn-gbm-clinical ID, Age_at_scan_years |
| Age SD | total age at the first available scan of each of the 630 subjects | 12.5 | tcia-upenn-gbm-clinical ID, Age_at_scan_years |
| Studies | contrast=T1w structural scans (T1; T1-Gd; T2; T2-FLAIR) in every session | 671 | bakas2022 Table 1 |
| Studies | contrast=T1w_ce | 671 | bakas2022 Table 1 |
| Studies | contrast=T2w | 671 | bakas2022 Table 1 |
| Studies | contrast=FLAIR | 671 | bakas2022 Table 1 |
| Studies | contrast=dwi DTI | 592 | bakas2022 Table 1 |
| Studies | contrast=dsc | 534 | bakas2022 Table 1 |
| Studies | contrast_set=dsc+dwi+FLAIR+T1w+T1w_ce+T2w | 527 | tcia-upenn-gbm-availability Structural imaging, DTI imaging, DSC imaging |
| Studies | contrast_set=FLAIR+T1w+T1w_ce+T2w | 72 | tcia-upenn-gbm-availability Structural imaging, DTI imaging, DSC imaging |
| Studies | contrast_set=dwi+FLAIR+T1w+T1w_ce+T2w | 65 | tcia-upenn-gbm-availability Structural imaging, DTI imaging, DSC imaging |
| Studies | contrast_set=dsc+FLAIR+T1w+T1w_ce+T2w | 7 | tcia-upenn-gbm-availability Structural imaging, DTI imaging, DSC imaging |
| Subjects | contrast=T1w | 630 | tcia-upenn-gbm-availability ID, Structural imaging |
| Subjects | contrast=T1w_ce | 630 | tcia-upenn-gbm-availability ID, Structural imaging |
| Subjects | contrast=T2w | 630 | tcia-upenn-gbm-availability ID, Structural imaging |
| Subjects | contrast=FLAIR | 630 | tcia-upenn-gbm-availability ID, Structural imaging |
| Subjects | contrast=dwi in at least one session | 554 | tcia-upenn-gbm-availability ID, DTI imaging |
| Subjects | contrast=dsc in at least one session | 497 | tcia-upenn-gbm-availability ID, DSC imaging |
| Subjects | contrast_set=dsc+dwi+FLAIR+T1w+T1w_ce+T2w contrasts pooled over a subject's sessions | 491 | tcia-upenn-gbm-availability ID, Structural imaging, DTI imaging, DSC imaging |
| Subjects | contrast_set=FLAIR+T1w+T1w_ce+T2w contrasts pooled over a subject's sessions | 70 | tcia-upenn-gbm-availability ID, Structural imaging, DTI imaging, DSC imaging |
| Subjects | contrast_set=dwi+FLAIR+T1w+T1w_ce+T2w contrasts pooled over a subject's sessions | 63 | tcia-upenn-gbm-availability ID, Structural imaging, DTI imaging, DSC imaging |
| Subjects | contrast_set=dsc+FLAIR+T1w+T1w_ce+T2w contrasts pooled over a subject's sessions | 6 | tcia-upenn-gbm-availability ID, Structural imaging, DTI imaging, DSC imaging |
| Studies | vendor=siemens | 663 | tcia-upenn-gbm-acquisition Manufacturer |
| Studies | vendor=ge | 8 | tcia-upenn-gbm-acquisition Manufacturer |
| Subjects | vendor=siemens | 622 | tcia-upenn-gbm-acquisition ID, Manufacturer |
| Subjects | vendor=ge | 8 | tcia-upenn-gbm-acquisition ID, Manufacturer |
| Studies | field_strength=3 one more session reports 2.89 T | 596 | tcia-upenn-gbm-acquisition Magnetic Field Strength |
| Studies | field_strength=1.5 | 74 | tcia-upenn-gbm-acquisition Magnetic Field Strength |
| Scans | total MR DICOM series | 3,680 | tcia-upenn-gbm Data Access (Version 2) |
| Images | total MR DICOM image files | 828,234 | tcia-upenn-gbm Data Access (Version 2) |
| Slides | total H&E whole-slide images in NDPI format | 71 | tcia-upenn-gbm Data Access (Version 2) |
| Slides | contrast=he | 71 | tcia-upenn-gbm Description |
Sources
The keys used in the table above.
- bakas2022 Bakas et al. 2022, Scientific Data 9, 453 paper
- tcia-upenn-gbm TCIA UPENN-GBM collection page (version 2, updated 2022-10-24) website
- tcia-upenn-gbm-clinical UPENN-GBM clinical table v2.1 (CC BY 4.0, open download), age at the first available scan per subject computed
- tcia-upenn-gbm-availability UPENN-GBM data availability per subject table (CC BY 4.0, open download) computed
- tcia-upenn-gbm-acquisition UPENN-GBM image acquisition parameters table (CC BY 4.0, open download) computed