BreastDCEDL_ISPY2 - DCE MRI dataset 982 cases

BreastDCEDL_ISPY2 Dataset (n=982) The BreastDCEDL_ISPY2 dataset is a curated subset of the I-SPY2 trial, comprising 982 breast cancer cases with pre-treatment DCE-MRI scans. It includes: DCE-MRI sequences (3–12 time points per scan, typically 7) Derived maps and full 3D tumor segmentations Clinical annotations: pathological complete response (pCR), hormone receptor (HR) status, HER2 status, MammaPrint risk level, and age at screening Demo: https://github.com/naomifridman/BreastDCEDL/blob/main/ISPY2/BrestDCEDL_ISPY2_demo.ipynb Citation @article{fridman2026breastdcedl, author = {Fridman, N. and Solway, B. and Fridman, T. and others}, title = {{BreastDCEDL}: A standardized deep learning-ready breast {DCE-MRI} dataset of 2,070 patients}, journal = {Scientific Data}, year = {2026}, doi = {10.1038/s41597-026-06589-6}, url = {https://doi.org/10.1038/s41597-026-06589-6}} Dataset Specifications Total Patients: 982 Total Size: ~54 GB Image Format: NIfTI (.nii.gz) Clinical Centers: 22+ institutions File Organization BreastDCEDL_ISPY2/ ├── BreastDCEDL_ISPY2_metadata.csv # Clinical and demographic data ├── dce/ # DCE-MRI sequences │ ├── ACRIN-6698-102212/ │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_0.nii.gz # Pre-contrast │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_1.nii.gz # Post-contrast 1 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_2.nii.gz # Post-contrast 2 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_3.nii.gz # Post-contrast 3 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_4.nii.gz # Post-contrast 4 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_5.nii.gz # Post-contrast 5 │ │ └── ACRIN-6698-102212_spy2_vis1_dce_aqc_6.nii.gz # Post-contrast 6 │ ├── ACRIN-6698-103939/ │ │ └── ... (3-12 DCE time points, typically 7) │ └── ... (982 patient directories total) └── masks/ # Tumor segmentations ├── ACRIN-6698-102212_spy2_vis1_mask.nii.gz ├── ACRIN-6698-103939_spy2_vis1_mask.nii.gz └── ... (982 binary mask files) Data Components Example of all acquisitions for random 7 patients, whith all the acquisitions. Resources Demo: https://github.com/naomifridman/BreastDCEDL/blob/main/ISPY2/BrestDCEDL_ISPY2_zenodo_demo.ipynb Full Methodology: Fridman et al., 2025 - arXiv:2506.12190 https://arxiv.org/abs/2506.12190 Citation @article{fridman2025breastdcedl, title={BreastDCEDL: A Comprehensive Breast Cancer DCE-MRI Dataset and Transformer Implementation for Treatment Response Prediction}, author={Fridman, Naomi and Solway, Bubby and Fridman, Tomer and Barnea, Itamar and Goldstein, Anat}, journal={arXiv preprint arXiv:2506.12190}, year={2025}, doi={10.48550/arXiv.2506.12190} }## BreastDCEDL_ISPY2 数据集（样本量n=982） 该数据集为I-SPY2试验的精选子集，纳入982例经治疗前动态对比增强磁共振成像（Dynamic Contrast-Enhanced Magnetic Resonance Imaging，DCE-MRI）扫描的乳腺癌病例，数据内容包含以下部分： 1. DCE-MRI序列：单次扫描包含3~12个时间点，常规为7个时间点； 2. 衍生影像图谱与完整的三维肿瘤分割掩码； 3. 临床标注信息：病理完全缓解（pathological complete response，pCR）、激素受体（hormone receptor，HR）状态、人表皮生长因子受体2（human epidermal growth factor receptor 2，HER2）状态、MammaPrint风险分级以及筛查时的年龄。 演示脚本：https://github.com/naomifridman/BreastDCEDL/blob/main/ISPY2/BrestDCEDL_ISPY2_demo.ipynb ## 引用文献 bibtex @article{fridman2026breastdcedl, author = {Fridman, N. and Solway, B. and Fridman, T. and others}, title = {{BreastDCEDL}: 面向深度学习的标准化乳腺癌DCE-MRI数据集（含2070例患者)}, journal = {Scientific Data}, year = {2026}, doi = {10.1038/s41597-026-06589-6}, url = {https://doi.org/10.1038/s41597-026-06589-6} } ## 数据集规格 - 总患者数：982例 - 总数据量：约54 GB - 影像存储格式：NIfTI（.nii.gz） - 临床参与机构：22家及以上医疗机构 ## 文件组织结构 BreastDCEDL_ISPY2/ ├── BreastDCEDL_ISPY2_metadata.csv # 临床与人口统计学数据表 ├── dce/ # DCE-MRI序列文件夹 │ ├── ACRIN-6698-102212/ │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_0.nii.gz # 预对比扫描图像 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_1.nii.gz # 首次对比增强后扫描图像 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_2.nii.gz # 第二次对比增强后扫描图像 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_3.nii.gz # 第三次对比增强后扫描图像 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_4.nii.gz # 第四次对比增强后扫描图像 │ │ ├── ACRIN-6698-102212_spy2_vis1_dce_aqc_5.nii.gz # 第五次对比增强后扫描图像 │ │ └── ACRIN-6698-102212_spy2_vis1_dce_aqc_6.nii.gz # 第六次对比增强后扫描图像 │ ├── ACRIN-6698-103939/ │ │ └── ... (3-12个DCE时间点，常规为7个) │ └── ... (总计982个患者专属文件夹) └── masks/ # 肿瘤分割掩码文件夹 ├── ACRIN-6698-102212_spy2_vis1_mask.nii.gz ├── ACRIN-6698-103939_spy2_vis1_mask.nii.gz └── ... (总计982个二值掩码文件) ## 数据组成 随机选取7例患者的完整采集序列示例。 ## 相关资源 1. 演示脚本：https://github.com/naomifridman/BreastDCEDL/blob/main/ISPY2/BrestDCEDL_ISPY2_zenodo_demo.ipynb 2. 完整方法学文献：Fridman等，2025年 - arXiv:2506.12190，链接：https://arxiv.org/abs/2506.12190 ## 引用文献 bibtex @article{fridman2025breastdcedl, title={{BreastDCEDL}: 乳腺癌DCE-MRI标准化数据集与治疗反应预测的Transformer实现}, author={Fridman, Naomi and Solway, Bubby and Fridman, Tomer and Barnea, Itamar and Goldstein, Anat}, journal={arXiv预印本 arXiv:2506.12190}, year={2025}, doi={10.48550/arXiv.2506.12190} }