Getting Started
Learn2Reg provides a common entry point to biomedical image registration benchmarks, evaluation resources, and scientific collaborations. Whether you are developing a registration method, reproducing published results, or participating in a challenge, the following steps will help you get started.
Explore the Benchmark Catalogue below to identify datasets and evaluation tasks relevant to your research. Each entry summarizes its scientific focus and provides links to the official resources.
Visit the official benchmark platform to review dataset availability, permitted uses, licensing conditions, evaluation protocols, and task-specific requirements. Some resources are publicly accessible, while others require registration or authorization from their custodians.
We encourage researchers to register their interest with the Learn2Reg Consortium. Registration helps us understand how resources are being used, support future benchmark development, and communicate citation and acknowledgment expectations. Registration may be required for consortium-managed resources or services where explicitly stated.
Access the resources through their official platforms, follow the applicable data-use and evaluation rules, and report the benchmark versions used. When publishing or publicly presenting results, cite the relevant Learn2Reg and dataset publications and acknowledge the Learn2Reg Consortium as described below.
Resource Use & Registration
The Learn2Reg Consortium promotes responsible use, transparent evaluation, and appropriate scientific attribution. We ask researchers using Learn2Reg resources to review the following expectations and register their intended use where applicable.
Researchers requesting consortium-managed resources or evaluation services should complete the registration process before access, when registration is required. Other researchers are welcome to register voluntarily to help us understand community needs and maintain communication.
The registration asks for your name, institution, contact information, intended resource, and a brief description of the proposed research. It also records acknowledgment of the applicable citation and responsible-use expectations.
A dedicated online registration form is being prepared. Until then, please email the consortium with your name, affiliation, resource of interest, intended research use, and confirmation that you have reviewed the citation and acknowledgment guidelines below.
Responsible Use Policy
Researchers using Learn2Reg resources are expected to:
- Follow the applicable dataset licenses, data-use agreements, privacy requirements, and challenge-specific rules.
- Use datasets and evaluation services only for purposes permitted by their respective custodians and platforms.
- Respect restrictions on redistributing datasets, annotations, test labels, or other controlled resources.
- Report the relevant dataset release, benchmark edition, preprocessing, evaluation split, and software version when publishing results.
- Provide appropriate citations and acknowledgments for the scientific resources used.
Important: The Learn2Reg Consortium connects and coordinates resources from multiple institutions. Registration with the consortium does not automatically grant access to restricted data or replace permissions required by individual custodians. Publicly available datasets and externally managed benchmark platforms remain governed by their published terms.
Citation & Acknowledgment
Scientific credit is essential to maintaining a sustainable and collaborative research community. We ask researchers using Learn2Reg benchmarks, datasets, or evaluation resources to recognize both the original scientific contributions and the community effort supporting these resources.
Citation Expectations
When publishing or publicly presenting research based on Learn2Reg resources, please:
- Cite Learn2Reg. Cite the foundational Learn2Reg publication when referring to the Learn2Reg benchmarking framework, together with the relevant challenge-edition publications where applicable.
- Cite the underlying resources. Include the required citations for the specific datasets, benchmarks, annotations, and software used in your research.
- Acknowledge the consortium. Acknowledge the Learn2Reg Consortium and reference this website when its resources or coordination contributed to the study.
We acknowledge the Learn2Reg Consortium (https://lab.dev.vanderbilt.edu/learn2reg/) for its contributions to the coordination and dissemination of biomedical image registration benchmark resources.
This acknowledgment supplements, rather than replaces, the scientific citations and attribution requirements of individual datasets and challenge organizers.
View Key Learn2Reg Publications
Foundational Learn2Reg Benchmark
Hering, A., Hansen, L., et al. “Learn2Reg: Comprehensive Multi-Task Medical Image Registration Challenge, Dataset and Evaluation in the Era of Deep Learning.” IEEE Transactions on Medical Imaging, 42(3), 697–712, 2023. DOI →
Learn2Reg 2024 Challenge
Hansen, L., Heyer, W., et al. “Learn2Reg 2024: New Benchmark Datasets Driving Progress on New Challenges.” Machine Learning for Biomedical Imaging, 2025. DOI →
LUMIR and Foundational Registration Models
Chen, J., Wei, S., et al. “Beyond the LUMIR challenge: The pathway to foundational registration models.” Medical Image Analysis, 113, 104175, 2026. DOI →
These are selected publications, not an exhaustive bibliography. Researchers should consult each benchmark’s official documentation for additional required references.
Benchmark Catalogue
Explore Learn2Reg challenges spanning medical imaging modalities, anatomical regions, and clinical applications. Each entry provides an overview of the scientific task and links to official benchmark resources.
Select a benchmark below to explore its details. Dataset availability, evaluation protocols, and submission conditions may vary by edition.
2026 — PSMAReg & Learn2Breath
The 2026 Learn2Reg Challenge features two complementary benchmarks addressing longitudinal whole-body oncology imaging and respiratory motion estimation.
PSMAReg
Longitudinal Whole-Body PSMA PET/CT Registration
PSMAReg addresses longitudinal registration of whole-body PSMA PET/CT examinations for therapy planning, dosimetry, and treatment response assessment. The benchmark focuses on aligning follow-up imaging with pre-therapy baseline scans despite anatomical changes, differences in imaging coverage, and treatment-related effects.
PSMAReg longitudinal registration example. Source: Official PSMAReg challenge.
Modalities: PSMA PET/CT | Anatomy: Whole body | Task: Longitudinal registration
The challenge describes 597 training/validation scans from 378 patients and 262 test scans from 131 patients. Evaluation emphasizes robust registration under anatomical changes while preserving quantitative PET information.
Explore PSMAReg — Task, Data & Results →
Learn2Breath
Inspiratory–Expiratory Chest CT Registration
Learn2Breath investigates respiratory motion through deformable registration of chest CT acquired at different breathing states. By aligning expiratory with inspiratory images, the benchmark supports research into lung biomechanics, regional deformation, and pulmonary disease.
Learn2Breath registration and deformation analysis example. Source: Official Learn2Breath challenge.
Modality: CT | Anatomy: Thorax | Task: Respiratory registration
The challenge describes 200 training pairs and 10 validation pairs, with a COPDGene test cohort selected for substantial respiratory volume changes. Evaluation focuses on accurate and anatomically plausible registration across breathing states.
Explore Learn2Breath — Task, Data & Submissions →
Dataset counts describe the 2026 challenge edition. Please consult the official competition pages for current rules, access instructions, and evaluation status. Official challenge descriptions →
2024–2025 — LUMIR: Brain MRI Registration
LUMIR focuses on unsupervised inter-subject brain MRI registration. Its benchmark resources include large-scale brain imaging data, while later evaluation tasks investigate robustness to domain shifts and multimodal image registration.
Representative brain MRI images from the LUMIR benchmark.
Modality: MRI | Anatomy: Brain | Focus: Inter-subject registration and generalization
2024–2025 — ReMIND2Reg: MRI to Intraoperative Ultrasound
ReMIND2Reg addresses registration between preoperative MRI and intraoperative ultrasound during brain tumor surgery. Based on a processed subset of ReMIND, the benchmark investigates cross-modal anatomical correspondence in the presence of changes occurring during surgical procedures.
Modalities: MRI and ultrasound | Anatomy: Brain | Focus: Image-guided surgery
2024 — COMULISglobe: Multimodal Microscopy
COMULISglobe extends Learn2Reg beyond radiological imaging to multimodal microscopy. It focuses on establishing spatial correspondence across complementary microscopy techniques, supporting the integration of biological information at microscopic scales.
Modalities: Microscopy | Focus: Multimodal registration
OncoReg — Registration for Clinical Oncology
OncoReg is a Learn2Reg-related initiative examining registration in clinical oncology, including transfer to radiotherapy data and other clinically motivated registration scenarios.
Focus: Oncology and clinically relevant image registration
2020–2023 — Earlier Learn2Reg Challenges
The earlier Learn2Reg challenges established a shared framework for evaluating registration algorithms across brain MRI, abdominal CT/MR, respiratory lung CT, and other anatomical and multimodal imaging tasks. These editions helped develop consistent approaches to evaluating accuracy, robustness, deformation plausibility, and computational efficiency.
Contribute a Dataset or Benchmark
Learn2Reg welcomes collaborations that expand the diversity, scientific relevance, and clinical impact of biomedical image registration research. Researchers and institutions interested in contributing datasets, annotations, new registration tasks, software, or evaluation expertise are encouraged to contact the organizing team.
Our long-term goal is to develop a connected catalogue of well-documented resources with transparent evaluation protocols, clear attribution, and appropriate access conditions.