voxel-wise unique co-variates

Dear all,

I am intersted to use 3D deconvolve to obtain voxel-wise response functions for event-related data in mice.

Particularly, for my analysis, I aim to investigate a modelled ‘nuisance’ term that has a unique shape for each individual brain voxel.

Essentially, I would like estimate the response function with and without this term, to evaluate if it would improve estimation of the response function.

I understand that in 3D deconvolve, it is possible to provide nuisance terms (such as motion vectors) in the baseline model.

I however don’t know how to provide such terms in a unique way for each respective voxel.

Can anyone please help me?

Thank you in advance,


Hi Michael,

3dREMLfit has a -dsort option that would apply a 4D dataset in such a way. Assuming alignment is not a problem, it could be passed by:

a. using afni_proc.py -copy_files to get both the HEAD and BRIK file:

-copy_files DSORT+orig.*

b. make sure 3dREMLfit runs, 3dDeconvolve does not, and passing that dataset for use in -dsort:

-regress_opts_reml -dsort DSORT+orig

I think that is enough to make it work.

Regarding alignment, this assumes that DSORT+orig is aligned with the EPI data at the time of running 3dREMLfit. That means DSORT+orig should already be aligned with the EPI volume registration base, and that you are not going to use afni_proc.py to align to the anat (anat to EPI is okay), or to use it to send the EPI to standard space. Again, the DSORT+orig dataset should be in alignment (and on the same grid) is the EPI data input to 3dREMLfit. Since you are working on mouse data, I expect this is a fair assumption.

Does that seem reasonable?

  • rick

Dear Rick,

Thank you so much for your reply. I will give it a try, your descriptions seem clear!