Single subject EPI-anat alignment fails (AFNI proc)

Dear all,

AFNI proc fails to properly align the EPI to the anatomical dataset for one specific subject. The resulting EPI brain by AFNI proc is completly misaligned.
Here is how I proceeded.

First, deoblique was required because the dataset was very oblique.

3dcopy $directory_raw/anat1to3d+orig Temporary
3drefit -oblique_recenter Temporary+orig
3drefit -deoblique Temporary+orig
3dcopy Temporary+orig Anat1_deoblique
rm Temporary+orig.BRIK
rm Temporary+orig.HEAD

Testing the overlap via the following code resulted in very good EPI-Anat overlaps for all subjects, except for subject 5 that is causing problems here.

@djunct_overlap_check \
    -ulay   Anat1_deoblique+orig \
    -olay   Rest2_deoblique+orig \
    -prefix img_olay_epi_anat

Second, I ran SSwarper without extra options. The SSwarper results for subject 5, which are of course forwarded to AFNI proc, are fine.

@SSwarper \
-input $directory_processed/Anat1_deoblique+orig \
-base MNI152_2009_template_SSW.nii.gz \
-subid $subject \
-odir $directory_sswarper_1

The image in the attachment of this post shows the EPI-anat overlap for subject 5 after the first deoblique code shown above. It is clear to see that the overlap is not good. However, adding -giant_move or even -ginormous_move in AFNI proc does not solve the problem for this specific subject. I tried out different cost functions beside lpc+ZZ too, such as lpa and nmi. This did also not help.

What further possibilities do I have to achieve a proper alignment for this subject?


Hi, Philipp-

This subject’s EPI appears to have very little structure and pretty large brightness inhomogeneities. That does pose a problem to alignment.

When the anatomical has been skullstripped, using “-cmass” should be pretty good at getting a good initial overlap, which is generally quite important for alignment. I think you might also want to make sure there is no additional skullstripping on this EPI; in, this would be done with:

-align_epi_strip_method  None

You miiiight also try:

-align_unifize_epi yes

… to help with some of the EPI inhomogeneity.