๐๐ฑ๐๐ข๐ญ๐๐ ๐ญ๐จ ๐ฌ๐ก๐๐ซ๐ ๐ญ๐ก๐ ๐ฉ๐ซ๐๐ฅ๐ข๐ฆ๐ข๐ง๐๐ซ๐ฒ ๐ซ๐๐ฅ๐๐๐ฌ๐ ๐จ๐ ๐๐จ๐ฌ๐๐ข๐๐๐๐.
To our knowledge, it is the ๐ฅ๐๐ซ๐ ๐๐ฌ๐ญ ๐๐ง๐ ๐ฆ๐จ๐ฌ๐ญ ๐๐ข๐ฏ๐๐ซ๐ฌ๐ ๐จ๐ฉ๐๐ง-๐ฌ๐จ๐ฎ๐ซ๐๐ ๐ซ๐๐ฐ ๐ฆ๐ฎ๐ฌ๐๐ฎ๐ฅ๐จ๐ฌ๐ค๐๐ฅ๐๐ญ๐๐ฅ ๐๐๐ ๐๐๐ญ๐๐ฌ๐๐ญ to date: 2,671 volumes, 80,156 slices, 454 patients, and 10 anatomies, with broad variation in contrast, orientation, and coil configuration.
This is a preliminary release and reflects only a fraction of the broader dataset we have been curating. We plan to expand MosaicMRI across institutions, field strengths, and new clinical challenges, with the goal of supporting both core ๐๐๐ ๐ฉ๐ซ๐จ๐๐ฅ๐๐ฆ๐ฌ (accelerated reconstruction, low-field MRI, motion compensation) and broader ๐๐ ๐๐ง๐ ๐๐จ๐ฎ๐ง๐๐๐ญ๐ข๐จ๐ง-๐ฆ๐จ๐๐๐ฅ ๐ช๐ฎ๐๐ฌ๐ญ๐ข๐จ๐ง๐ฌ in science and healthcare, including robustness, generalization, scaling laws, heterogeneous data mixtures, continual learning, and test-time adaptation.
๐๐๐๐ฌ๐ข๐ญ๐:
mosaicmri.ai/
๐๐๐ฉ๐๐ซ:
arxiv.org/abs/2604.11762
๐๐จ๐๐:
github.com/AIF4S/mosaicmri
๐๐๐๐๐ฌ๐ฌ:
mosaicmri.ai/#access
๐๐๐ง๐๐ก๐ฆ๐๐ซ๐ค:
mosaicmri.ai/benchmark/
๐๐๐ซ๐ฅ๐ฒ ๐๐ฏ๐๐ฅ๐ฎ๐๐ญ๐ข๐จ๐ง ๐ญ๐ซ๐๐๐ค๐ฌ ๐๐ซ๐ ๐ง๐จ๐ฐ ๐ฅ๐ข๐ฏ๐. Official benchmark tracks and community challenges will follow.
Grateful to NIH National Library of Medicine (NLM) for supporting this project.