generate `sample_mask`

[pbellec]

generate `sample_mask`

[pbellec]

re behavior of ICA-AROMA and compcor: good catch.
It would indeed be important to incorporate an example of that in the example.
So many moving parts in this project!!

Re scrubbing and despiking, indeed in Ciric 2017 scrubbing is implemented through removing time points while despiking is done by adding one-hot regressors. But I think this is a bit arbitrary: nothing prevents to implement scrubbing with one-hot regressors.
I am now thinking it may make more sense to have two separate loading methods load and load_mask, depending on whether or not the user plans on using sample_mask. See my comments on this PR: #142 I don't have a strong opinion at this stage, I am just trying to find a way for users to understand what to do with the mask, and forcing an explicit API call may help. Please let me know what you think.

[htwangtw]

There are some overlapping elements of three of the remaining issues before the beta release, I will address some related issues with the following plan:

1. refactor current code base non-steady-state regressors (in fMRIprep confound output) and modify motion scrubbing method generated regressors to sample_mask_ (issue #132 and #26)
2. Get new test data with rmsd and non_steady_state_outlierXX (issue #127 and #26)
3. Write the despiking regressor method; possibly reuse some element of the existing scrubbing method (issue #127)
   Step one will need the most inputs and discussions. Let me know what you think.

[htwangtw]

I thought I finished working on this but realise we need to double-check if ICA-AROMA nifti has non-steady-state volumes trimmed or not. ICA-AROMA and compcor based methods all trimmed out non-steady-state volumes when calculating the metrics. Needs to check on the new test data I am processing.