EPI to MNI space alignment issues

Hey there,

I’m having some alignment issues when I transform my epi data from orig space to MNI space. I preprocessed about 50 subjects worth of data through the proc script and about 7 of them appear to have the same alignment issues. The epi aligns no problem to the skull stripped anatomy in orig space, and the anatomy seems to transform to MNI space with no problems. However, when the mat.warp.aff12.1D matrix is applied to the tshifted epi dataset in the 3dAllineate step the resulting volreg dataset is markedly misaligned to the anatomy.

I’ve tried several things:

  1. In 3dAllineate I’ve tried various cost functions, changed the -mast_dxyz options, tried the -automask option, and various -final options. Nothing helped.
  2. I moved to 3dQwarp with the -resample option and while this fixed the sizing component of the alignment issue there are some interesting signal smears outside of the brain. I tried 3dQwarp with -allineate sample and that was a mess.
  3. I then tried @Align_Centers to align the tshifted data to the stripped anatomy in MNI space and reran the 3dQwarp with -resample and this got rid of the smears outside of the brain but there is still substantial signal drop out inside the brain.

Do you have any other ideas for how to improve the alignment?

I’ve attached screenshots of the alignment output from the proc script as well as what the alignment looks like currently from the @Align_centers/3dQwarp output.

Thanks in advance!


How are you generating these transformations - afni_proc.py, align_epi_anat.py, 3dAllineate, …?

I started out using -afni_proc.py and when the alignment failed I looked specifically at options I could change in 3dAllineate on its own. Once those didn’t work out then I moved to the 3dQwarp command and tried various options there.

My guess is that you’ve got the directions or order of transformations wrong here. If you would like, you can upload some data. I’ll PM some instructions.