FSL FEAT 2nd level analysis

Following the last SCT 6.1 Course, I made a pipeline for lumbar fMRI acquisitions with one subject and two different contrast acquisitions in the subject space.
I registered the second acquisition with SCT in the subject space of the first one.

Then, I ran each FSL Feat 1st level analysis, deactivating the registration step to the template.

But FSL Feat can’t run for the second level analysis because he needs to be fed with the registration at the first level.

Do you have an idea of how to bypass this issue?

Hi Charles,

For 2nd level (group) analysis you de facto need to get all your images in the same space. Similarly to brain fMRI analysis, you can use the subject-wise transformation that brings the fMRI derived 1st level analysis (eg z-score) to the template space. If you use a software like SPM or FSL, you need to make sure the warping field is compatible with these software, and for that you can use sct_image as instructed here.

Related post:

Hi Charles,

I also came across this issue when doing a subject level analysis over different first levels (in my case runs) and had already brought them into the same space but FSL did not recognize it. I found this video by Jeanette Mumford super helpful in that regard:

Here she explains how to trick FSL Feat into recognizing that you did some registration.
In short you need to do a fake (I think minimum 6 DOF) registration to some image. Then it creates ‘reg’ folders in your first level folder structure. Then you just replace the example_func2standard.mat with the identity matrix (can be found in your FSL directory under etc/flirtsch/ident.mat )
and replace standard.nii.gz with an image with the same dimensions as your first level images. I took for example a mean image of my different runs but I think it does not matter much. That way no additional registration step is done to your images but you can proceed with the analysis in Feat.

Maybe that helps?


Yes Ulrike absolutely correct. FSL needs to have a reg folder to proceed with 2nd level analysis. It will read all the transformation/warp files in reg folder and create a new “reg_standard” folder to save all cope/varcope files in standard space. You need to make sure all necessary files in reg folder so better copy a full reg folder and just replace some files in it. I usually try to avoid using FEAT for the 2nd level analysis - I normally warp all cope images into standard space, merge them all together and use “randomise” directly for group analysis. Hope this helps.


:open_mouth: amazing !!! It is exactly what I wanna do! Thank you so much !!!
It’s super helpful.