BraTS documentation¶

BraTS Orchestrator¶

______         _____ _____
| ___ \       |_   _/  ___|
| |_/ /_ __ __ _| | \ `--.
| ___ \ '__/ _` | |  `--. \
| |_/ / | | (_| | | /\__/ /
\____/|_|  \__,_\_/ \____/


 _____          _               _             _
|  _  |        | |             | |           | |
| | | |_ __ ___| |__   ___  ___| |_ _ __ __ _| |_ ___  _ __
| | | | '__/ __| '_ \ / _ \/ __| __| '__/ _` | __/ _ \| '__|
\ \_/ / | | (__| | | |  __/\__ \ |_| | | (_| | || (_) | |
 \___/|_|  \___|_| |_|\___||___/\__|_|  \__,_|\__\___/|_|

Providing the top-performing algorithms from the Brain Tumor Segmentation (BraTS) challenges, through an easy-to-use Python API powered by Docker.

Features¶

Access to top-performing algorithms from recent BraTS challenges
Easy-to-use minimal API
Extensive documentation and examples

Installation¶

With a Python 3.8+ environment, you can install BraTS orchestrator directly from PyPI:

# lightweight base package
pip install brats
# With preprocessing functionalities
pip install brats[preprocessing]

Important

To run BraTS orchestrator, you require a Docker installation.
Many algorithms also require GPU support (NVIDIA Docker).
In case you do not have access to a CUDA-capable GPU, the overview tables in the Available Algorithms and Usage section indicate which algorithms are CPU compatible.

Docker and NVIDIA Container Toolkit Setup¶

Docker: Installation instructions on the official website
NVIDIA Container Toolkit: Refer to the NVIDIA install guide and the official GitHub page

Singularity Support¶

BraTS orchestrator also supports Singularity as an alternative to Docker. To enable Singularity, install it following the official guide and specify the backend to use as Backends.SINGULARITY when running the inference:

from brats.constants import Backends
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    output_file="path/to/segmentation.nii.gz",
    backend=Backends.SINGULARITY
)

Available Algorithms and Usage¶

Important

BraTS challenge algorithms require preprocessed brain images. See section Data preprocessing requirements

Segmentation Challenges¶

Note: Some legacy segmentation algorithms from BraTS challenges before 2023 are available via BraTS Toolkit.

Adult Glioma Segmentation (Pre & Post-Treatment)¶

Adult Glioma Segmentation on pre- and post-treatment brain MRI exams.

Usage example (code) and top 3 participants

from brats import AdultGliomaPreAndPostTreatmentSegmenter
from brats.constants import AdultGliomaPreAndPostTreatmentAlgorithms

segmenter = AdultGliomaPreAndPostTreatmentSegmenter(algorithm=AdultGliomaPreAndPostTreatmentAlgorithms.BraTS25_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Note: If you’re interested in Adult Glioma Segmentation, the BrainLes GlioMODA package may also be of interest.

Class: brats.AdultGliomaPreAndPostTreatmentSegmenter (Docs)
Challenge Paper 2023: Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Ishika Jain, et al.	N/A	❌	BraTS25_1
2025	2nd	Qu Lin, et al.	N/A	✅	BraTS25_2
2025	3rd	Liwei Jin, et al.	N/A	✅	BraTS25_3A
2025	3rd	Adrian Celaya, et al.	N/A	❌	BraTS25_3B
2024	1st	André Ferreira, et al.	Link	❌	BraTS24_1
2024	2nd	Heejong Kim, et al.	Link	❌	BraTS24_2
2024	3rd	Adrian Celaya	N/A	✅	BraTS24_3

Note: The MNI152 atlas, available on Zenodo, was employed for registration in the 2024 and subsequent BraTS Glioma Post-treatment Segmentation challenges.

Adult Glioma Segmentation (Pre-Treatment)¶

Adult Glioma Segmentation on pre-treatment brain MRI exams.

Usage example (code) and top 3 participants

from brats import AdultGliomaPreTreatmentSegmenter
from brats.constants import AdultGliomaPreTreatmentAlgorithms

segmenter = AdultGliomaPreTreatmentSegmenter(algorithm=AdultGliomaPreTreatmentAlgorithms.BraTS23_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Note: If you’re interested in Adult Glioma Segmentation, the BrainLes GlioMODA package may also be of interest.

Class: brats.AdultGliomaPreTreatmentSegmenter (Docs)
Challenge Paper 2023: Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2023	1st	André Ferreira, et al.	Link	❌	BraTS23_1
2023	2nd	Andriy Myronenko, et al.	Link	❌	BraTS23_2
2023	3rd	Fadillah Adamsyah Maani, et al.	Link	❌	BraTS23_3

Note: The SRI24 atlas, available on Zenodo, was employed for registration in the 2023 and prior BraTS Glioma Pre-Treatment Segmentation challenges.

BraTS-Africa Segmentation¶

Adult Glioma Segmentation on brain MRI exams in Sub-Sahara-Africa patient population.

Usage example (code) and top 3 participants

from brats import AfricaSegmenter
from brats.constants import AfricaAlgorithms

segmenter = AfricaSegmenter(algorithm=AfricaAlgorithms.BraTS25_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Class: brats.AfricaSegmenter (Docs)
Challenge Paper 2023 Link
Challenge Paper 2024: N/A

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Claudia Takyi Ankomah, et al.	N/A	✅	BraTS25_1
2025	2nd	William Boonzaier, et al.	N/A	❌	BraTS25_2A
2025	2nd	Mohtady Barakat, et al.	N/A	❌	BraTS25_2B
2025	3rd	Ahmed Jaheen, et al.	N/A	❌	BraTS25_3
2024	1st	Abhijeet Parida, et al.	Link	❌	BraTS24_1
2024	2nd	Yanguang Zhao, et al.	Link	✅	BraTS24_2
2024	3rd	Sarim Hashmi, et al.	Link	❌	BraTS24_3
2023	1st	Andriy Myronenko, et al.	Link	❌	BraTS23_1
2023	2nd	Alyssa R Amod, et al.	Link	❌	BraTS23_2
2023	3rd	Ziyan Huang, et al.	Link	✅	BraTS23_3

Note: The SRI24 atlas, available on Zenodo, was employed for registration in BraTS Africa Segmentation challenges.

Meningioma Segmentation¶

Segmentation of Meningioma on brain MRI exams.

Usage example (code) and top 3 participants

from brats import MeningiomaSegmenter
from brats.constants import MeningiomaAlgorithms

### Example for 2023 algorithms
segmenter = MeningiomaSegmenter(algorithm=MeningiomaAlgorithms.BraTS25_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation_23.nii.gz",
)

Class: brats.MeningiomaSegmenter (Docs)
Challenge Paper 2023 Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Yu Haitao, et al.	N/A	❌	BraTS25_1
2025	2nd	Mohammad Mahdi Danesh Pajouh, et al.	N/A	❌	BraTS25_2
2023	1st	Andriy Myronenko, et al.	Link	❌	BraTS23_1
2023	2nd	Ziyan Huang, et al.	Link	✅	BraTS23_2
2023	3rd	Daniel Capellán-Martín et al.	Link	❌	BraTS23_3

Meningioma Radio Therapy Segmentation¶

Segmentation of Meningioma on T1C brain MRI exams.

Usage example (code) and top 3 participants

from brats import MeningiomaRTSegmenter
from brats.constants import MeningiomaRTAlgorithms

segmenter = MeningiomaRTSegmenter(algorithm=MeningiomaRTAlgorithms.BraTS25_1, cuda_devices="0")
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    output_file="segmentation_24.nii.gz",
)

Class: brats.MeningiomaRTSegmenter (Docs)
Challenge Paper 2024 Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Valeria Abramova, et al.	N/A	❌	BraTS25_1
2025	2nd	Sanskriti Srivastava, et al.	N/A	✅	BraTS25_2
2025	3rd	Nima Sadeghzadeh, et al.	N/A	✅	BraTS25_3
2024	1st	Valeria Abramova	N/A	❌	BraTS24_1
2024	2nd	Mehdi Astaraki	N/A	❌	BraTS24_2
2024	3rd	Andre Ferreira, et al.	Link	✅	BraTS24_3

Note: The MRI dataset in the Meningioma-Radiotherapy challenge was provided in native space. However, the SRI24 atlas, available on Zenodo, was employed for registration in BraTS Meningioma Pre-operative challenges.

Brain Metastases Segmentation¶

Segmentation on brain metastases on MRI exams for pre- and post-treatment cases.

Usage example (code) and top 3 participants

from brats import MetastasesSegmenter
from brats.constants import MetastasesAlgorithms

segmenter = MetastasesSegmenter(algorithm=MetastasesAlgorithms.BraTS25_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Note: If you’re interested in Brain Metastases Segmentation, the BrainLes AURORA package may also be of interest.

Class: brats.MetastasesSegmenter (Docs)
Challenge Paper 2023 Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Maria Bancerek, et al.	N/A	❌	BraTS25_1
2025	2nd	Wes Krikorian, et al.	N/A	✅	BraTS25_2
2023	1st	Andriy Myronenko, et al.	Link	❌	BraTS23_1
2023	2nd	Siwei Yang, et al.	Link	❌	BraTS23_2
2023	3rd	Ziyan Huang, et al.	Link	✅	BraTS23_3

Note: The SRI24 atlas, available on Zenodo, was employed for registration in BraTS Metastasis segmentation challenges.

Pediatric Segmentation¶

Segmentation of pediatric brain tumors on MRI exams.

Usage example (code) and top 3 participants

from brats import PediatricSegmenter
from brats.constants import PediatricAlgorithms

segmenter = PediatricSegmenter(algorithm=PediatricAlgorithms.BraTS25_1, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Note: If you’re interested in Pediatric Segmentation, the BrainLes PeTu package may also be of interest.

Class: brats.PediatricSegmenter (Docs)
Challenge Paper 2024 Link
Challenge Results Paper 2023 Link
Challenge Paper 2023 Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Yuxiao Yi, et al.	N/A	❌	BraTS25_1
2025	2nd	Meng-Yuan Chen, et al.	N/A	❌	BraTS25_2
2025	3rd	Haitao Yu, et al.	N/A	❌	BraTS25_3
2024	1st	Mehdi Astaraki	N/A	❌	BraTS24_1
2024	2nd	Tim Mulvany, et al.	Link	❌	BraTS24_2
2024	3rd	Sarim Hashmi, et al.	Link	❌	BraTS24_3
2023	1st	Daniel Capellán-Martín, et al.	Link	❌	BraTS23_1
2023	2nd	Andriy Myronenko, et al.	Link	❌	BraTS23_2
2023	3rd	Yubo Zhou	Link	❌	BraTS23_3

Note: The SRI24 atlas, available on Zenodo, was employed for registration in BraTS Pediatric Tumor Segmentation challenges.

Generalizability Across Tumors (BraTS-GoAT) Segmentation¶

Segmentation algorithm, adapting and generalizing to different brain tumors with segmentation labels of different tumor sub-regions.

Usage example (code) and top 3 participants

from brats import GoATSegmenter
from brats.constants import GoATAlgorithms

segmenter = GoATSegmenter(algorithm=GoATAlgorithms.BraTS25_1A, cuda_devices="0")
# these parameters are optional, by default the latest winning algorithm will be used on cuda:0
segmenter.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="segmentation.nii.gz",
)

Class: brats.PediatricSegmenter (Docs)
Challenge Paper 2024: N/A

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Meng-Yuan Chen, et al.	N/A	❌	BraTS25_1A
2025	1st	To-Liang Hsu, et al.	N/A	❌	BraTS25_1B
2025	1st	Vaidehi Satushe, et al.	N/A	❌	BraTS25_1C
2025	1st	Simone Bendazzoli, et al.	N/A	❌	BraTS25_1D
2024	1st	Frank Miao, Shengjie Niu	N/A	❌	BraTS24_1

Note: The datasets used in this challenge were adapted from other segmentation challenges, so the atlas type depends on the original dataset.

Inpainting Challenge¶

Algorithm to realistically synthesize and fill 3D healthy brain tissue in a region affected by glioma in brain MRI exams.

Usage example (code) and top 3 participants

from brats import Inpainter
from brats.constants import InpaintingAlgorithms

inpainter = Inpainter(algorithm=InpaintingAlgorithms.BraTS25_1A, cuda_devices="0")
inpainter.infer_single(
    t1n="path/to/voided_t1n.nii.gz",
    mask="path/to/mask.nii.gz",
    output_file="inpainting.nii.gz",
)

Class: brats.Inpainter (Docs)
Challenge Paper 2023 and 2024 Link

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	Juexin Zhang, et al.	N/A	❌	BraTS25_1A
2025	1st	André Ferreira, et al.	N/A	❌	BraTS25_1B
2025	2nd	Juhyung Ha, et al.	N/A	❌	BraTS25_2
2024	1st	Juexin Zhang et al.	Link	✅	BraTS24_1
2024	2nd	André Ferreira, et al.	Link	❌	BraTS24_2
2024	3rd	Alicia Durrer, et al.	N/A	❌	BraTS24_3
2023	1st	Juexin Zhang, et al.	Link	✅	BraTS23_1
2023	2nd	Alicia Durrer, et al.	Link	❌	BraTS23_2
2023	3rd	Jiayu Huo, et al.	Link	✅	BraTS23_3

Note: The datasets used in this challenge were adapted from other segmentation challenges, so the atlas type depends on the original dataset.

Missing MRI Challenge¶

Algorithm to realistically synthesize missing MRI modalities from available sequences to enhance brain tumor segmentation.

Usage example (code) and top 3 participants

from brats import MissingMRI
from brats.constants import MissingMRIAlgorithms

missing_mri = MissingMRI(algorithm=MissingMRIAlgorithms.BraTS25_1, cuda_devices="0")
# Example to synthesize t2f modality (whichever modality is missing will be inferred)
missing_mri.infer_single(
    t1c="path/to/t1c.nii.gz",
    t1n="path/to/t1n.nii.gz",
    # t2f="path/to/t2f.nii.gz",
    t2w="path/to/t2w.nii.gz",
    output_file="inferred_t2f.nii.gz",
)

Class: brats.MissingMRI (Docs)
Challenge Paper 2024: N/A

Year	Rank	Author	Paper	CPU Support	Key Enum
2025	1st	André Ferreira, et al.	N/A	❌	BraTS25_1
2025	2nd	Agustin Ujarky Cartaya Lathulerie, et al.	N/A	❌	BraTS25_2
2025	3rd	Lina Chator, et al.	N/A	✅	BraTS25_3
2024	1st	Jihoon Cho et al.	Link	❌	BraTS24_1
2024	2nd	Haowen Pang	N/A	❌	BraTS24_2
2024	3rd	Minjoo Lim et al.	Link	❌	BraTS24_3
2023	1st	Ivo Baltruschat	Link	❌	BraTS23_1

Note: The datasets used in this challenge were adapted from other segmentation challenges, so the atlas type depends on the original dataset.

Tip

For a full notebook example with more details, please check here:

Data Preprocessing Requirements¶

BraTS challenge algorithms require preprocessed brain scans. This preprocessing typically includes co-registration, brain extraction, and registration to a challenge-specific brain atlas (template). Please refer to each challenge’s documentation for details on which template to use.

For convenience, we provide challenge-specific wrappers around our preprocessing package, part of the BrainLesion Suite. You can install the package with the preprocessing extra using:

pip install brats[preprocessing]

The modular architecture of the BraTS Orchestrator facilitates employing your own preprocessing routines. Alternatively, the preprocessing package can be used independently to design custom preprocessing pipelines tailored to your specific needs.

Citation¶

Important

If you use BraTS orchestrator in your research, please cite it to support the development!

Kofler, F., Rosier, M., Astaraki, M., Baid, U., Möller, H., Buchner, J. A., Steinbauer, F., Oswald, E., Rosa, E. de la, Ezhov, I., See, C. von, Kirschke, J., Schmick, A., Pati, S., Linardos, A., Pitarch, C., Adap, S., Rudie, J., Verdier, M. C. de, … Menze, B. (2025). BraTS orchestrator: Democratizing and Disseminating state-of-the-art brain tumor image analysis arXiv preprint arXiv:2506.13807

@misc{kofler2025bratsorchestratordemocratizing,
      title={BraTS orchestrator : Democratizing and Disseminating state-of-the-art brain tumor image analysis},
      author={Florian Kofler and Marcel Rosier and Mehdi Astaraki and Ujjwal Baid and Hendrik Möller and Josef A. Buchner and Felix Steinbauer and Eva Oswald and Ezequiel de la Rosa and Ivan Ezhov and Constantin von See and Jan Kirschke and Anton Schmick and Sarthak Pati and Akis Linardos and Carla Pitarch and Sanyukta Adap and Jeffrey Rudie and Maria Correia de Verdier and Rachit Saluja and Evan Calabrese and Dominic LaBella and Mariam Aboian and Ahmed W. Moawad and Nazanin Maleki and Udunna Anazodo and Maruf Adewole and Marius George Linguraru and Anahita Fathi Kazerooni and Zhifan Jiang and Gian Marco Conte and Hongwei Li and Juan Eugenio Iglesias and Spyridon Bakas and Benedikt Wiestler and Marie Piraud and Bjoern Menze},
      year={2025},
      eprint={2506.13807},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2506.13807},
}

Contributing¶

We welcome all kinds of contributions from the community!

Reporting Bugs, Feature Requests, and Questions¶

Please open a new issue here.

Code Contributions¶

Nice to have you on board! Please have a look at our CONTRIBUTING.md file.

Sections¶

Getting Started:

Readme

API Reference: