MAMA-MIA

MAMA-MIA: A large-scale multicenter breast cancer DCE-MRI benchmark dataset with expert segmentations

Welcome to the MAMA-MIA dataset repository! This dataset contains 1,506 cases of breast cancer dynamic contrast-enhanced magnetic resonance images (DCE-MRI) with expert tumor segmentations. Below, you will find all the necessary information to download and use the dataset, as well as instructions on how to run inference using our pre-trained nnUNet model.

📢 Join the MAMA-MIA Challenge!

Are you working with breast cancer DCE-MRI data? Put your algorithms to the test in the MAMA-MIA Challenge – an international benchmark on tumour segmentation and treatment response prediction!

🔬 2 Tasks · 🧠 Real-world data · 🏆 Evaluation on private multicenter datasets

🚀 Submit your models to Codabench 🔗 Participate now on Codabench

The MAMA-MIA Challenge paper is now available on IEEE Transactions on Medical Imaging: 🔗 https://ieeexplore.ieee.org/document/11690473

If you participate in the challenge or use its framework, please cite the arXiv paper.

If you use the MAMA-MIA dataset or the pretrained model in your research, please cite our publication and the dataset publications of the images included in the dataset.

BibTex:

@article{garrucho2025,
  title={A large-scale multicenter breast cancer DCE-MRI benchmark dataset with expert segmentations},
  author={Garrucho, Lidia and Kushibar, Kaisar and Reidel, Claire-Anne and Joshi, Smriti and Osuala, Richard and Tsirikoglou, Apostolia and Bobowicz, Maciej and Riego, Javier del and Catanese, Alessandro and Gwoździewicz, Katarzyna and Cosaka, Maria-Laura and Abo-Elhoda, Pasant M and Tantawy, Sara W and Sakrana, Shorouq S and Shawky-Abdelfatah, Norhan O and Salem, Amr Muhammad Abdo and Kozana, Androniki and Divjak, Eugen and Ivanac, Gordana and Nikiforaki, Katerina and Klontzas, Michail E and García-Dosdá, Rosa and Gulsun-Akpinar, Meltem and Lafcı, Oğuz and Mann, Ritse and Martín-Isla, Carlos and Prior, Fred and Marias, Kostas and Starmans, Martijn P A and Strand, Fredrik and Díaz, Oliver and Igual, Laura and Lekadir, Karim},
  journal = {Scientific Data},
  year = {2025},
  doi = {10.1038/s41597-025-04707-4},
  pages = {453},
  number = {1},
  volume = {12}
}

@article{mamamiachallenge,
  author={Garrucho, Lidia and Joshi, Smriti and Kushibar, Kaisar and Osuala, Richard and Bobowicz, Maciej and Bargalló, Xavier and Jaruševičius, Paulius and Geissler, Kai and Schäfer, Raphael and Alberb, Muhammad and Xu, Tony and Martel, Anne and Sleiman, Daniel and Awasthi, Navchetan and Awwad, Hadeel and Vilanova, Joan C. and Martí, Robert and Schouten, Daan and Lee, Jeong Hoon and Rusu, Mirabela and Poeta, Eleonora and Vargas, Luisa and Pastor, Eliana and Zuluaga, Maria A. and Kächele, Jessica and Bounias, Dimitrios and Ertl, Alexandra and Gwoździewicz, Katarzyna and Cosaka, Maria-Laura and Abo-Elhoda, Pasant M. and Tantawy, Sara W. and Sakrana, Shorouq S. and Shawky-Abdelfatah, Norhan O. and Abdo-Salem, Amr Muhammad and Kozana, Androniki and Divjak, Eugen and Ivanac, Gordana and Nikiforaki, Katerina and Klontzas, Michail E. and García-Dosdá, Rosa and Gulsun-Akpinar, Meltem and Lafcı, Oğuz and Martín-Isla, Carlos and Díaz, Oliver and Igual, Laura and Lekadir, Karim},
  journal={IEEE Transactions on Medical Imaging}, 
  title={The MAMA-MIA Challenge: Advancing Generalizability and Fairness in Breast MRI Tumor Segmentation and Treatment Response Prediction}, 
  year={2026},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TMI.2026.3733343}}

Potential Applications

Some of the potential applications of the dataset are:

• Classification, Detection and Segmentation of Breast Cancer in MRI using Deep Learning.
• Radiomic extraction and Machine Learning.
• Segmentation Model Finetuning using the Pretrained nnUNet Weigths.
• Fine-tuning of Foundational Models.
• Treatment Response to Neoadjuvant Chemotherapy.
• Patient Survival Prediction.
• Segmentation Quality Control.
• Image Synthesis.

Repository Structure

├── README.md                     # Project overview, setup instructions, and usage details
├── requirements.txt              # List of Python package dependencies for running the project
├── nnUNet                        # Forked nnUNet GitHub repository for training the automatic segmentation model
├── notebooks                      # Jupyter Notebooks for project demos and examples
│   ├── demo_preprocessing_pipeline.ipynb   # Demonstrates common MRI preprocessing steps (e.g., bias field correction, normalization, resampling)
│   ├── display_clinical_data.ipynb         # Example on how to parse and visualize clinical data from the dataset
│   └── display_segmentations.ipynb         # Various methods for displaying MRI images and their corresponding segmentation masks
└── src                             # Python source code used in the Jupyter Notebooks
    ├── clinical_data.py           # Functions for handling and processing clinical data
    ├── preprocessing.py           # Preprocessing functions
    └── visualization.py           # Functions for visualizing images, segmentations, and metrics

Downloading the Dataset

The MAMA-MIA dataset is hosted on Synapse. You can download the dataset using the CLI or Python with the following code:

Log in using a Personal Access Token (recommended):

• Create a Personal Access Token

Command Line Interface (CLI)

 synapse login --authToken <YOUR_TOKEN_HERE>
 synapse get -r syn60868042

Python

import synapseclient
syn = synapseclient.Synapse()
syn.login(authToken="YOUR_TOKEN_HERE")  # or syn.login('username', 'password')
entity = syn.get("syn60868042")

Check Synapse documentation for more info.

Alternative Download (XNAT)

If you are unable to access Synapse due to regional or institutional restrictions, the dataset is also available through the Health-RI XNAT platform:

Download MAMA-MIA dataset from XNAT

You will might to create an XNAT account (if you do not already have one) to access to the MAMA-MIA project before downloading the data.

Content

Expert Segmentations

Running Inference with nnUNet pre-trained model

The pre-trained vanilla nnUNet model has been trained using the 1506 full-image DCE-MRIs and the expert segmentations from the MAMA-MIA dataset.

Full-image Tumor Segmentation	Dice	IoU	HD95	MSD
nnUNet Fold 0	0.7600	0.6512	39.90	11.89
nnUNet Fold 1	0.7617	0.6500	41.25	12.10
nnUNet Fold 2	0.7749	0.6714	31.12	8.86
nnUNet Fold 3	0.7616	0.6577	36.25	11.74
nnUNet Fold 4	0.7520	0.6396	38.57	10,81
nnUNet Mean	0.7620	0.6539	37.41	11.08

The metrics correspond to Dice Coefficient, Intersection over Union (IoU), the 95 Percentile of Hausdorff Distance (HD95) and to the Mean Surface Distance (MSD). The distances are in mm. The library used to compute these metrics was seg-metrics v1.2.7.

Step 1. Clone the repository

Clone the forked repository of nnUNet.

git clone https://github.com/LidiaGarrucho/MAMA-MIA
cd MAMA-MIA/nnUNet

Step 2. Install the necessary dependencies

To run the pre-trained nnUNet model, follow the installation instructions.

Step 3. Download the pre-trained weights

The nnUNet pretrained weights can be dowloaded from Synapse. Unzip the folder inside nnUNet GitHub repository under nnUNet/nnunetv2/nnUNet_results.

Step 4. (Recommended) Preprocess your input MRI images

The recommended preprocessing steps to get optimum performance are:

• z-score normalization. For DCE-MRI, use the mean and standard deviation of all the phases (from pre to last post-contrast) to z-score the DCE-MRI sequence.
• isotropic pixel spacing. The MRIs were resampled using a uniform pixel spacing of [1,1,1].

Step 5. Run the nnUNet inference

nnUNetv2_predict -i /path/to/your/images -o /path/to/output -d 101 -c 3d_fullres

• Replace /path/to/your/images with the directory containing your input images.
• Replace /path/to/output with the directory where you want to save the output segmentations.

Note: An error might arise if your images are not in compressed NifTI format (.nii.gz).

Citation and Attribution

If you use the MAMA-MIA dataset or the pretrained model in your research, please cite our publication and the dataset publications of the images included in the dataset.

MAMA-MIA Dataset

Garrucho, L., Kushibar, K., Reidel, CA. et al. A large-scale multicenter breast cancer DCE-MRI benchmark dataset with expert segmentations. Sci Data 12, 453 (2025) (10.1038/s41597-025-04707-4).

BibTex:

@ARTICLE{mamamiachallenge,
  author={Garrucho, Lidia and Joshi, Smriti and Kushibar, Kaisar and Osuala, Richard and Bobowicz, Maciej and Bargalló, Xavier and Jaruševičius, Paulius and Geissler, Kai and Schäfer, Raphael and Alberb, Muhammad and Xu, Tony and Martel, Anne and Sleiman, Daniel and Awasthi, Navchetan and Awwad, Hadeel and Vilanova, Joan C. and Martí, Robert and Schouten, Daan and Lee, Jeong Hoon and Rusu, Mirabela and Poeta, Eleonora and Vargas, Luisa and Pastor, Eliana and Zuluaga, Maria A. and Kächele, Jessica and Bounias, Dimitrios and Ertl, Alexandra and Gwoździewicz, Katarzyna and Cosaka, Maria-Laura and Abo-Elhoda, Pasant M. and Tantawy, Sara W. and Sakrana, Shorouq S. and Shawky-Abdelfatah, Norhan O. and Abdo-Salem, Amr Muhammad and Kozana, Androniki and Divjak, Eugen and Ivanac, Gordana and Nikiforaki, Katerina and Klontzas, Michail E. and García-Dosdá, Rosa and Gulsun-Akpinar, Meltem and Lafcı, Oğuz and Martín-Isla, Carlos and Díaz, Oliver and Igual, Laura and Lekadir, Karim},
  journal={IEEE Transactions on Medical Imaging}, 
  title={The MAMA-MIA Challenge: Advancing Generalizability and Fairness in Breast MRI Tumor Segmentation and Treatment Response Prediction}, 
  year={2026},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TMI.2026.3733343}}

The MAMA-MIA dataset includes public DCE-MRI images from four different collection in the TCIA repository under the following licenses:

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

Breast-MRI-NACT-Pilot (License CC BY 3.0)

Newitt, D., & Hylton, N. (2016). Single site breast DCE-MRI data and segmentations from patients undergoing neoadjuvant chemotherapy (Version 3) [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/K9/TCIA.2016.QHsyhJKy

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

Thank you for using our dataset and pretrained model! If you have any questions or issues, please feel free to open an issue in this repository.