OpenBTAI Brain Metastases
A comprehensive dataset of annotated brain metastasis MR images with clinical and radiomic data
637 longitudinal brain MRI studies of 75 deceased patients with 260 brain metastases from Spanish hospitals, with 593 semi-automatic segmentations on post-contrast T1w, clinical and treatment data, and morphological and radiomic features.
Overview
This collection from the Mathematical Oncology Laboratory at the University of Castilla-La Mancha gathers the follow-up brain MRI of patients with brain metastases, collected under the retrospective multicenter OpenBTAI study. Each segmented exam comes with lesion masks, and the release adds spreadsheets of clinical and treatment history, morphological measurements and PyRadiomics features. It targets metastasis detection and segmentation, response assessment, separating tumor progression from radiation necrosis, and survival modeling. All files on Figshare are released under CC0.
Composition
75 adult patients, all deceased, contribute 637 imaging studies covering 260 metastases. The primary cancers were non-small cell lung cancer (38), small cell lung cancer (5), breast cancer (22), melanoma (6), ovarian cancer (2), kidney cancer (1) and uterine cancer (1). Every patient has high-resolution post-contrast T1-weighted imaging, and most studies also hold other series such as T1-weighted, T2-weighted, FLAIR and diffusion images. The clinical spreadsheet lists 47 women and 28 men.
Acquisition
Patients were diagnosed between 2005 and 2021 at Spanish hospitals and scanned on GE, Philips and Siemens scanners at 1, 1.5 or 3 tesla, mostly 1.5 tesla. Post-contrast T1w series had to have pixel spacing and slice thickness of at most 2 mm with no slice gap. Raw series are shared as DICOM in six archives; segmented series and masks are shared as NIfTI in the original image space. Dates are shifted so that the first metastasis scan of each patient falls on 1 January 1900, and images are defaced.
Annotations
593 post-contrast T1w series were segmented with an in-house threshold tool, corrected slice by slice by one researcher, cross-checked by experienced researchers and corrected by a radiologist. Each lesion carries two labels: an enhancing part and a non-enhancing or necrotic part. Radiation necrosis was confirmed for 39 lesions. Lesion centroids in MNI space are provided.
Known limitations
- Only patients who had died were included, which skews the cohort toward advanced disease.
- Counts per sequence other than post-contrast T1w are not reported.
- The GPA prognostic score is available only for some institutions.
- Masks cover 154 distinct metastases, not all 260.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 75 subjects.
Sex
- Female 47 63%
- Male 28 37%
Contrast / sequence
scans, values can overlap
- T1-weighted, contrast enhanced 593
Condition
subjects, values can overlap
- Brain metastases 75 100%
- Lung cancer 43 57%
- Non-small cell lung cancer 38 51%
- Breast cancer 22 29%
- Melanoma 6 8%
- Kidney cancer 1 1%
Scanner vendor
scans
- GE HealthCare 225
- Philips 197
- Siemens Healthineers 171
Field strength
scans
- 1.5 T 550
- 3 T 35
- 1 T 8
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
Creative Commons Zero 1.0 Universal
Public domain dedication. Do anything with the data, including commercial use, without asking and without having to give credit.
What you can do
- Yes
- Yes
- Yes
- Yes
What you can share
- Yes
- Yes
- Yes
What you must do
- No
- Share alike No
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Citation
Ocaña-Tienda B, Pérez-Beteta J, Villanueva-García JD, et al. A comprehensive dataset of annotated brain metastasis MR images with clinical and radiomic data. Scientific Data 10, 208 (2023). https://doi.org/10.1038/s41597-023-02123-0
All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Subjects | total | 75 | ocanatienda2023 Methods: Subject characteristics |
| Studies | total Longitudinal follow-up MRI studies | 637 | ocanatienda2023 Methods: Subject characteristics |
| Scans | contrast=T1w_ce Segmented post-contrast T1w sequences only | 593 | ocanatienda2023 Methods: Image acquisition |
| Scans | field_strength=1 Of the 593 segmented post-contrast T1w sequences | 8 | ocanatienda2023 Methods: Image acquisition |
| Scans | field_strength=1.5 Of the 593 segmented post-contrast T1w sequences | 550 | ocanatienda2023 Methods: Image acquisition |
| Scans | field_strength=3 Of the 593 segmented post-contrast T1w sequences | 35 | ocanatienda2023 Methods: Image acquisition |
| Scans | vendor=ge Of the 593 segmented post-contrast T1w sequences | 225 | ocanatienda2023 Methods: Image acquisition |
| Scans | vendor=philips Of the 593 segmented post-contrast T1w sequences | 197 | ocanatienda2023 Methods: Image acquisition |
| Scans | vendor=siemens Of the 593 segmented post-contrast T1w sequences | 171 | ocanatienda2023 Methods: Image acquisition |
| Subjects | condition=brain_metastasis | 75 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | condition=nsclc Primary tumor | 38 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | condition=lung_cancer Primary tumor; sum of NSCLC (38) and SCLC (5) | 43 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | condition=breast_cancer Primary tumor | 22 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | condition=melanoma Primary tumor | 6 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | condition=kidney_cancer Primary tumor | 1 | ocanatienda2023 Methods: Subject characteristics |
| Subjects | sex=female | 47 | openbtai-clinical Sex column (1 = female) |
| Subjects | sex=male | 28 | openbtai-clinical Sex column (2 = male) |
Sources
The keys used in the table above.
- ocanatienda2023 Ocaña-Tienda et al. 2023, Scientific Data paper
- figshare-openbtai-mets Figshare collection with file listing and per-item license website
- openbtai-clinical OpenBTAI_METS_ClinicalData.xlsx (CC0), counted per unique patient id computed