Request access to MosaicMRI
Manual review is required before download access is granted. Please use your real research identity and the same account details you use for MosaicMRI requests.
MosaicMRI is released for non-commercial research and educational use only under the USC Data Sharing Agreement. Review the full terms at https://www.mosaicmri.ai/data-sharing-agreement.html before submitting your request.
Log in or Sign Up to review the conditions and access this dataset content.
MosaicMRI
Overview
MosaicMRI is a large-scale raw musculoskeletal MRI dataset with manual-approval gated access for non-commercial research and educational use.
MosaicMRI is a diverse dataset of fully sampled raw musculoskeletal MRI designed for reconstruction, robustness, and generalization research. The project materials describe 10 anatomies, 2,671 volumes, and 80,156 slices spanning realistic variation in anatomy, contrast, orientation, and coil configuration.
What is in this repository
This repository contains chunked tar archives for the MosaicMRI release:
- Multicoil train split
- Multicoil validation split
- Multicoil test split
- Anatomy transfer challenge data
- Contrast generalization challenge data
Current upload count: 43 data files.
Access
- Create or sign in to your Hugging Face account.
- Open the gated access form on this dataset page.
- Provide your institution, intended research use, and agreement acknowledgement.
License and use restrictions
This dataset is distributed under a custom research-only agreement:
- License name:
usc-dsa-v1 - Full terms: https://www.mosaicmri.ai/data-sharing-agreement.html
MosaicMRI is for internal, non-commercial research and educational use only.
Benchmark and project information
The broader MosaicMRI project includes benchmark tasks focused on raw MRI reconstruction and generalization. Documentation, project context, and citation details are available on the website:
- Website: https://www.mosaicmri.ai
- Data sharing agreement: https://www.mosaicmri.ai/data-sharing-agreement.html
Citation
Please cite:
Arguello, P., Tinaz, B., Mohammad, S. S., Soltanolkotabi, Maryam, and Soltanolkotabi, Mahdi. MosaicMRI: A Diverse Dataset and Benchmark for Raw Musculoskeletal MRI. arXiv (2026). https://arxiv.org/abs/2604.11762
Repository
adminmosaic/MosaicMRI
- Downloads last month
- 32