search

neuropoly / idea-projects

Ideas for cool projects
1 stars 0 forks source link

Automated tool to detect artefacted image (can use the 'exclude.yaml' files) #22

Open jcohenadad opened 8 months ago

jcohenadad commented 8 months ago

As we are building our internal MRI database, we progressively list the images that have excessive artifacts in the file exclude.yaml. We should exploit that information and train a model that automatically detects if an images has excessive artifact. When processing large cohorts, that information would be very useful to have.

Todo: make sure that our exclude.yaml uses the same vocabulary to identify artifacts.

plbenveniste commented 8 months ago

That's a very nice and useful idea. With the numerous problematic images in CanProCo, I think it can easily be trained (at least on the PSIR contrast : examples here).

valosekj commented 8 months ago

Great idea! Cross-referring https://github.com/neuropoly/idea-projects/issues/5.