Datasets:

Bubenpo
/

BreastDividerDataset

language:
  - en
pretty_name: Breast Divider
tags:
  - Breast
  - MRI
  - 3d
license: cc-by-nc-sa-4.0
task_categories:
  - image-segmentation
  - image-classification
size_categories:
  - 10K<n<100K

BreastDividerV2: A Large-Scale Dataset and Model for Left–Right Breast MRI Segmentation and Lesion Analysis

This is v2.0.0 of our dataset! Look for v1.0.0 under tags for the original publication.

[MICCAI 2025 WOMEN] BreastDivider: A Large-Scale Dataset and Model for Left–Right Breast MRI Segmentation

Read the paper:

Authors: Maximilian Rokuss*, Benjamin Hamm*, Yannick Kirchhoff*, Klaus Maier-Hein
*equal contribution

🧠 Introduction

Breast MRI plays a pivotal role in breast cancer detection, diagnosis, and treatment planning. However, most existing segmentation models fail to distinguish between the left and right breasts, limiting their usefulness in tasks such as unilateral classification, response evaluation, or post-mastectomy follow-up.

In this work, we introduce the first publicly available large-scale dataset with explicit left and right breast segmentation labels, comprising over 17,000 3D MRI scans. Accompanying this dataset is a robust nnU-Net–based segmentation model, trained specifically to identify and separate left and right breast regions in clinical MRI data. This resource provides a foundation for developing anatomically aware AI models and enables large-scale pretraining for downstream breast MRI tasks. Additionally, in V2 we provide over 3,000 classification targets for healthy, benign, and malignant breasts, as well as 467 segmentation masks.

📂 Dataset

This repository includes a complete 17k+ MRI scan dataset with left/right segmentation masks, over 3000 lesion classification targets and 467 lesion segmentation masks. We curated a diverse and comprehensive breast MRI dataset by aggregating scans from multiple publicly available sources:

Duke-Breast-Cancer-MRI dataset
MAMA-MIA
Advanced-MRI-Breast-Lesions
EA1141
ODELIA
ISPY1
ISPY2

In total, the dataset comprises 17,956 3D scans. It includes a variety of common MRI modalities such as:

T1-weighted (T1)
T1 with contrast (T1+C)
T2-weighted (T2)
FLAIR
Diffusion-weighted imaging (DWI)

The Advanced-MRI-Breast-Lesions dataset contains T1 DCE with multiple fat-saturated phases, delayed T1, T2, and T1 non-fat saturated sequences.
The Duke dataset includes pre-operative DCE MRI at 1.5T or 3T, with fat-saturated and non-fat saturated sequences.
The EA1141 dataset, collected across 48 clinical sites, features non-contrast and post-contrast T1-weighted, T2-weighted, and DWI sequences.
The ODELIA dataset includes images across 5 clinical sites containing T2-weighted acquisition alongside a DCE acquisition with one pre-contrast phase and between 2 and 7 post-contrast phases per study.

To ensure consistency and quality, we only included MRI volumes with:

At least 32 slices per axis
A resolution of ≤ 3×3×3 mm

📂 Dataset Folder Structure

dataset/
├── imagesTr_batch1/
│   ├── BreastDivider_00001_0000.nii.gz
│   ├── ...
│   └── BreastDivider_10000_0000.nii.gz
├── imagesTr_batch2/
│   ├── BreastDivider_10001_0000.nii.gz
│   ├── ...
│   └── BreastDivider_17956_0000.nii.gz
├── labelsTr_batch1/
│   ├── BreastDivider_00001.nii.gz
│   ├── ...
│   └── BreastDivider_10000.nii.gz
├── labelsTr_batch2/
│   ├── BreastDivider_10001.nii.gz
│   ├── ...
│   └── BreastDivider_17956.nii.gz
├── lesion_annotations/
│   ├── classification/
│   │   └── classification.csv
│   ├── segmentation/
│   │   ├── DUKE_001.nii.gz
│   │   ├── ...
│   │   └── ISPY2_982505.nii.gz
│   ├── dwi_channel_mapping.json
│   └── dwi_channel_mapping_explanation.txt
├── dataset.json
└── breastdivider_id_mapping.csv

imagesTr_batch1 & imagesTr_batch2: Training images in .nii.gz format (split into two batches).
labelsTr_batch1 & labelsTr_batch2: Corresponding breast label masks in .nii.gz format (split into two batches).
lesion_annotations/classification: Contains classification.csv with 3021 lesion classification targets.
lesion_annotations/segmentation: Contains 467 lesion segmentation masks for bilateral images from the Mama Mia dataset.
lesion_annotations/dwi_channel_mapping.json contains the mapping for breast lesion classification and breast lesion segmentation, indicating which DWI MRI phase image corresponds to which image in the dataset.
dataset.json: nnU-Net style dataset metadata
breastdivider_id_mapping.csv: Mapping between original dataset IDs and internal BreastDivider IDs

📄 Citation

If you use this dataset or model in your work, please cite:

@article{rokuss2025breastdivider,
  title     = {Divide and Conquer: A Large-Scale Dataset and Model for Left–Right Breast MRI Segmentation},
  author    = {Rokuss, Maximilian and Hamm, Benjamin and Kirchhoff, Yannick and Maier-Hein, Klaus},
  journal   = {arXiv preprint arXiv:2507.13830},
  year      = {2025}
}

The Breast Divider dataset includes public DCE-MRI images from several different collections under the following licenses:

EA1141 (License CC BY-NC 4.0)

Comstock, C. E., Gatsonis, C., Newstead, G. M., Snyder, B. S., Gareen, I. F., Bergin, J. T., Rahbar, H., Sung, J. S., Jacobs, C., Harvey, J. A., Nicholson, M. H., Ward, R. C., Holt, J., Prather, A., Miller, K. D., Schnall, M. D., & Kuhl, C. K. (2023). Abbreviated Breast MRI and Digital Tomosynthesis Mammography in Screening Women With Dense Breasts (EA1141) (Version 1) [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/2BAS-HR33

AMBL (License CC BY-NC 4.0)

Daniels, D., Last, D., Cohen, K., Mardor, Y., & Sklair-Levy, M. (2024). Standard and Delayed Contrast‑Enhanced MRI of Malignant and Benign Breast Lesions with Histological and Clinical Supporting Data (AMBL) (Version 2) [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/C7X1-YN57

ISPY1 Trial (License CC BY 3.0)

David Newitt, Nola Hylton, on behalf of the I-SPY 1 Network and ACRIN 6657 Trial Team. (2016). Multi-center breast DCE-MRI data and segmentations from patients in the I-SPY 1/ACRIN 6657 trials. The Cancer Imaging Archive. https://doi.org/10.7937/K9/TCIA.2016.HdHpgJLK

ISPY2 Trial (License CC BY 4.0)

Li, W., Newitt, D. C., Gibbs, J., Wilmes, L. J., Jones, E. F., Arasu, V. A., Strand, F., Onishi, N., Nguyen, A. A.-T., Kornak, J., Joe, B. N., Price, E. R., Ojeda-Fournier, H., Eghtedari, M., Zamora, K. W., Woodard, S. A., Umphrey, H., Bernreuter, W., Nelson, M., … Hylton, N. M. (2022). I-SPY 2 Breast Dynamic Contrast Enhanced MRI Trial (ISPY2) (Version 1) [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/TCIA.D8Z0-9T85

Duke-Breast-Cancer-MRI (License CC BY-NC 4.0)

Saha, A., Harowicz, M. R., Grimm, L. J., Weng, J., Cain, E. H., Kim, C. E., Ghate, S. V., Walsh, R., & Mazurowski, M. A. (2021). Dynamic contrast-enhanced magnetic resonance images of breast cancer patients with tumor locations [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/TCIA.e3sv-re93

MAMA-MIA (License CC BY-NC 4.0)

Garrucho, L., Kushibar, K., Reidel, C.-A., Joshi, S., Osuala, R., Tsirikoglou, A., Bobowicz, M., del Riego, J., Catanese, A., Gwoździewicz, K., Cosaka, M.-L., Abo-Elhoda, P. M., Tantawy, S. W., Sakrana, S. S., Shawky-Abdelfatah, N. O., Salem, A. M. A., Kozana, A., Divjak, E., Ivanac, G., Nikiforaki, K., Klontzas, M. E., García-Dosdá, R., Gulsun-Akpinar, M., Lafcı, O., Mann, R., Martín-Isla, C., Prior, F., Marias, K., Starmans, M. P. A., Strand, F., Díaz, O., Igual, L., & Lekadir, K. (2025). A large-scale multicenter breast cancer DCE-MRI benchmark dataset with expert segmentations. https://doi.org/10.1038/s41597-025-04707-4

ODELIA (License CC BY-NC 4.0)

Müller-Franzes, G., Escudero Sánchez, L., Payne, N., Athanasiou, A., Kalogeropoulos, M., Lopez, A., Soro Busto, A. M., Camps Herrero, J., Rasoolzadeh, N., Zhang, T., Mann, R., Jutz, D., Bode, M., Kuhl, C., Veldhuis, W., Saldanha, O. L., Zhu, J., Kather, J. N., Truhn, D., & Gilbert, F. J., on behalf of the ODELIA Consortium. (2025). A European Multi-Center Breast Cancer MRI Dataset. https://doi.org/10.48550/arXiv.2506.00474