search

ME-ICA / tedana

TE-dependent analysis of multi-echo fMRI
https://tedana.readthedocs.io
GNU Lesser General Public License v2.1
161 stars 95 forks source link

Curate annotated ICA results for decision tree validation #652

Open tsalo opened 3 years ago

tsalo commented 3 years ago

Summary

In order to facilitate validation of the tedana decision tree, we can curate a set of ICA outputs

Additional Detail

Perhaps we should look into using SwipesForScience to crowdsource annotations.

I can also transfer this issue to the tedana-comparison repo if necessary.

Next Steps

  1. Determine the approach we want to use.
  2. Collect a few runs from a range of multi-echo datasets.
  3. Minimally preprocess each run (STC+MC).
  4. Run tedana on each run, and retain the mixing matrix and T2* map (and possibly component table?) from each. These inputs can be used to completely reproduce a given tedana run.
  5. Generate component-wise reports (without classifications).
  6. Annotate the components, whether via crowdsourcing or using a couple of experienced annotators.
tsalo commented 3 years ago

@\richford is working on a SwipesForScience cookiecutter that could make this much easier to do. Once that is set up, we might have a real chance for crowdsourcing classifications!