BreasTomo-Synth: An in silico-generated Digital Breast Tomosynthesis dataset for AI-based tumor segmentation
Authors/Creators
- 1. Universidad de Tarapacá
-
2.
Pontificia Universidad Católica de Chile
Contributors
- 1. Universidad de Tarapacá
-
2.
Pontificia Universidad Católica de Chile
- 3. Millennium Institute for Intelligent Healthcare Engineering, Santiago, Chile
Description
BreastTomoSynth is an in silico-generated Digital Breast Tomosynthesis (DBT) dataset containing regions of interest (ROI) of spiculated tumours and healthy breast tissue. The tiff format files include in-silico-generated breast tumor images and their respective segmentation masks. Each ROI is of size 109x109 pixels. The dataset was created with VICTRE, an FDA-cleared software from Badano et al. [1]
Future dataset versions will include the complete DBT images and files with the insertion coordinates of each simulated tumor.
Notes
We have now expandend the dataset.
The release contains two complementary components:
ROI dataset: 174 two-dimensional regions of interest (ROIs), each 109 × 109 pixels, with corresponding binary segmentation masks. The dataset includes both lesion-present and lesion-absent ROIs.
DBT dataset: 30 reconstructed DBT image stacks, each distributed as a 15-slice multi-page TIFF file, with a corresponding binary segmentation mask stack.
All segmentation masks were manually delineated and reviewed by an experienced medical imaging professional. Mask values are 0 for background and 255 for lesion.
The dataset also includes separate metadata files for the ROI and DBT components.
Files
DBT_dataset.zip
Files
(145.1 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:d2a8be4c7580aab16b42e2aebd47ab83
|
142.9 MB | Preview Download |
|
md5:fece7bcf08f7502669e98b3855eaacf7
|
1.8 kB | Preview Download |
|
md5:4817174487fe3c29f5fe9345d9c66632
|
4.7 kB | Preview Download |
|
md5:415e47eb2aa6d9118c313e1b427a0e2d
|
2.1 MB | Preview Download |
|
md5:a4ca89d18a4c84275638e4cc9aff42a3
|
7.2 kB | Preview Download |
Additional details
Additional titles
- Alternative title
- Synthetic digital breast tomosynthesis dataset with breast tumor segmentation masks
Related works
- References
- Journal article: 10.1001/jamanetworkopen.2018.5474 (DOI)
Funding
- Agencia Nacional de Investigación y Desarrollo
- Beca de Doctorado Nacional 21212497
Dates
- Available
-
2024-12-04v1
- Available
-
2026-08-22v1.1
Software
- Repository URL
- https://github.com/nacca-sudo/syntdbt
- Programming language
- Python