A voxel-level lesion Digital Breast Tomosynthesis segmentation dataset derived from the publicly available Breast Cancer Screening Digital Breast Tomosynthesis (BCS-DBT) dataset

The dataset consists of 201 patients with 396 DBT images, including MLO and CC views with biopsy confirmation. The dataset is intended to support a broad range of applications, including lesion segmentation, radiomics, and the development of deep learning models for DBT image analysis. A more detailed description and annotation protocol can be found at the preprint doi.org/10.21203/rs.3.rs-10669063/v1.

This repo provides processed DBT images, lesion segmentation files and structured BI-RADS information (label.zip). Additionally, we provided two trained segmentation nnUNet weights (model_weights.zip) for further validation using the MIC-DKFZ/nnUNet code. This work was supported by grants from Marie Skłodowska-Curie Doctoral Networks Actions (HORIZON-MSCA-2021-DN-01-01; 101073222) , Cancerfonden (22-2389 Pj) and Natural Science Foundation of Fujian Province, China (2024J011643, to Muzhen He).

The computations were enabled by the Berzelius resource provided by the Knut and Alice Wallenberg Foundation at the National Academic Infrastructure for Supercomputing in Sweden. We also acknowledge the Cancer Imaging Archive team and BSC-DBT dataset owner for making the imaging and clinical data used in this study publicly available.

Citation

Zhikai Yang, Qingxia Zheng, Jacob Schultzén, Zexuan Chen, Örjan Smedby, Muzhen He, Rodrigo Moreno BCSDBT-Seg Lesion Segmentation Annotations and Structured BI-RADS Reports for Breast Tomosynthesis Dataset, 11 August 2026, PREPRINT (Version 1) available at Research Square [doi.org/10.21203/rs.3.rs-10669063/v1]