NARPS Open Pipelines Results

[bclenet]

Goal

• Create a dataset containing the reproduced results from the NARPS Open Pipelines project.
• Find a way to archive the correlation values for these results as well.

Why this is useful

This would help the community to work on these results without having to rerun the pipelines.

Timeline

No response

More specifically

• The dataset could be a datalad sub-dataset.
• Data could be hosted on Neurovault.
• Results for subsets of subjects (20, 40, 60, 80) could be added as well.
• Provenance should be tracked for these results (at least the version of the project used to generate them, and the software environment).
• Do not archive subject level results (it would not be compliant with GDPR), archive group level results only.
• The dataset should conform to BIDS.
• Result files do not need to be versioned for now, meaning each result for a new NARPS Open Pipelines version erases the previous file.

Open questions

[alexpron]

Thanks for the issue @bclenet , here are a few thoughts:

• data/reproduced should be a BIDS as much as possible compliant derivatives
  a gross structure according to https://bids-website.readthedocs.io/en/latest/getting_started/folders_and_files/derivatives.html could be :

  reproduced
    preprocessing
         team-XXXX
    processing (l1 analysis ?)
    group (l2, analysis)
           team-XXXX 
               nubjects=XXXX
  

                           
-  only the group level should be shared (group level should be a datalad dataset, if we want to also versioned results we could
-  results do not need to be versioned (see either to launch using datalad-containers run or 
-  remote of the  group level datalad dataset should be on neurovault 
- when to allow datalad push on remote ?

    + manually ?
    + automatically for us if success of the checks and specific datalad tag present
 
    

[alexpron]

On neurovault, NARPS results are shared as one collection per pipeline (see https://neurovault.org/collections/4876/)
There are already some tests :

• https://neurovault.org/collections/15001/
• https://neurovault.org/collections/10383/ (reproduction)
  correlation maps could be also added directly into it ?