Organized by researchers from @aicentre.dk @ucph.bsky.social @mghmartinos.bsky.social @harvardmed.bsky.social @bispebjerghospital.bsky.social
@stefanocerri.bsky.social @asbjornmunk.bsky.social @peirong26.bsky.social
@fomochallenge.bsky.social
Foundation Model Challenge for Brain MRI hosted @ #MICCAI2025! π§
Organized by researchers from @aicentre.dk @ucph.bsky.social @mghmartinos.bsky.social @harvardmed.bsky.social @bispebjerghospital.bsky.social
@stefanocerri.bsky.social @asbjornmunk.bsky.social @peirong26.bsky.social
π°π· Join us for a half-day workshop in South Korea #MICCAI2025!
Read more about the challenge and signup on:
β‘οΈ fomo25.github.io
π€: huggingface.co/datasets/FOMO25/FOMO-MRI
π»: github.com/fomo25/baseline-codebase
π» Comprehensive Code-Release
Participants are provided with a baseline framework to carry out both pretraining and finetuning, enabling them to concentrate on specific components of the workflow.
π₯ Few-shot Evaluation on Clinical Data
Both tracks will be evaluated on three few-shot and out-of-domain tasks on clinical data:Β infarct detection, meningioma segmentation, andΒ brain age estimation.
No restrictions on fine-tuning method.
π Two submission tracks:Β "Methods" and "Open"
1οΈβ£: Train a model onΒ FOMO-60K and competebest pretraining method.
2οΈβ£: Train a foundation model on any combination of data (both public and private) and showcase your foundation model with no restrictions.
πΏ Large-scale Dataset
We release FOMO-60K, a large-scale dataset of 60,529 brain MRI from diverse set of resources, including clinical scans!
Dataset has been skull-stripped, co-registered and RAS-oriented, making it straightforward to start pretraining on.
Available on π€
FOMO25 accepted at #MICCAI2025! π
Excited to announce the first challenge at MICCAI focusing on the development of self-supervised pretraining of foundation models for brain MRI! π§
With access to a large-scale dataset, codebase, cash prices and multiple tracks.
Read more β¬οΈ