08MQ reproduction #121
08MQ reproduction #121
Conversation
Thank you! |
There was a problem hiding this comment.
@bclenet: I've reviewd the pipeline code as suggested and did not spot any issues. One thing that would help me doing a second check would be to get the corresponding fsf file and open it into FSL. Is this feasible with nipype? (I think the equivalent is possible for SPM but I am unsure this is available for FSL).
I have also answered in this review your 3 questions, when can continue the discussion directly in the comments.
Let me know if there is any other check I can help with!
narps_open/pipelines/team_08MQ.py
Outdated
| # FLAMEO Node - Estimate model | ||
| estimate_model = Node(FLAMEO(), name = 'estimate_model') | ||
| estimate_model.inputs.run_mode = 'fe' # Fixed effect | ||
| estimate_model.inputs.mask_file = Info.standard_image('MNI152_T1_2mm_brain_mask.nii.gz') |
There was a problem hiding this comment.
This should be a mask created for each subject (as each subject might have different voxels missing in the MNI space due to signal dropouts). You can use the intersection of the masks generated by the run-level GLM. Does this help?
There was a problem hiding this comment.
Ok thanks, let's do this !
There was a problem hiding this comment.
Discussed on Dec 1. It looks like feat is creating the mask in the background we can check in the paper that reproduce feat in nipype how they dealt with this.
narps_open/pipelines/team_08MQ.py
Outdated
| randomise.inputs.vox_p_values = True | ||
| randomise.inputs.c_thresh = 0.05 | ||
| randomise.inputs.tfce_E = 0.01 | ||
| randomise.inputs.mask = Info.standard_image('MNI152_T1_2mm_brain_mask.nii.gz') |
There was a problem hiding this comment.
Is it mandatory to specify a mask? If not we can leave this empty so that randomise comupte the mask itself. If we have to provide a mask, for the same reason as above (i.e. some voxels might be missing because of signal-dropouts), we need to use here a study-specific mask. I think that typically we use the intersection of all the subject-specific masks (but maybe this is something to check -- I can look into it, let me know!)
There was a problem hiding this comment.
It is not mandatory to provide a mask, I can remove the line then.
There was a problem hiding this comment.
Erratum : the FLAMEO node (couple of lines earlier) in the group level analysis uses a mandatory mask input : I guess we will have to compute the intersection of all previously computed subject-specific masks then.
There was a problem hiding this comment.
Let's use the same mask for randomize as well.
|
I can get fsf files for the run level analysis (as soon as it runs...) : it is the output of the |
…ost function [skip ci]
This Pull Request is related to issue #120
Changes proposed in this Pull Request:
Checklist:
mainbranch