I’m hoping for some input about the best way to go about fMRIPrep, thinking long-term about reporting results from analyses.
First off, thanks for releasing the boilerplate with a CC0 license for reuse. That makes life infinitely easier.
In my case, my thought is to run fMRIPrep with as many options as possible to give myself the richest density of outputs for future analyses. If I do so, I have the following questions:
Lets assume I get AROMA confounds and fsLR/cifti outputs in addition to MNI152 outputs but I don’t use them in an analysis. Is there a useful way to regenerate the boilerplate with only the things I did use (e.g., just MNI152 space and non-aroma confound estimation).
I have resting state and task-based fMRI. Is there an advantage to running fMRIPrep without the --task-id flag. If so, is there a recommended way of explaining it, in the context of the boilerplate indicating that preprocessing was done across all tasks/sessions/runs, to justify the inclusion of non-resting state in the pipeline but not reporting any findings?
Just trying to think ahead to some future manuscripts. Thanks.
There’s the --boilerplate option that will only generate the boilerplate text, so you can use whatever options you want and regenerate it.
The advantage to running fMRIPrep without the --task-id flag is just to preprocess all task data without submitting multiple jobs. The advantage to running it with is mostly for parallelizing (assuming that the anatomical workflow is already complete).
Not yet, but we’d be happy to include some if you have a suggestion. I can see a couple options, depending on the journal’s attitude to pre-generated text.
If they would prefer it to be exactly what is generated, then having a statement, such as “The above text applies to the full range of steps performed by fMRIPrep. For the analysis presented in this manuscript, only tasks A, B and C are relevant.”
Or they may prefer an edited boilerplate and a statement summarizing the changes made. “The above text was modified from the boilerplate text automatically generated by fMRIPrep to exclude irrelevant tasks and resampling targets.”
Also, noting that it is CC0, you are free to modify the text. Our objection (purely theoretical at this point, AFAIK) is to journals requiring authors to modify it, but it’s not our goal to restrict authors’ discretion to improve the accuracy of the text.