Hello AFNI-gurus!
I’m aligning EPI to anat using align_epi_anat.py. Here is the command I’m using:
align_epi_anat.py -anat {Part[{SLURM_ID}]}_T1_S+orig
-epi {Let}_{run}_Cat_S_TS_VR+orig.
-volreg off
-epi_base 0
-deoblique off
-tshift off
-epi2anat
-save_tsh
-save_vr
-cost lpc
-giant_move
-suffix _A+orig
-epi_strip None
-anat_has_skull no
-overwrite
But after this step I’m getting stray voxels outside the brain (Aligned.png) that did not exist before alignment. This is happening in every subject, across all runs,. And I don’t have any concerning movement (as seen from volreg, which I’m running separately) so the ghosting is likely not due to movement.
Additionally, normalising this data cause a lot of bleeding out of the brain(Normalised.png). That’s concerning because when I run 3dDeconvolve on this data I’m getting (the much-elusive, corrected) “blobs” outside the brain. So I’m a little hesitant to ignore the problem! Does anyone have any suggestions about why this might be happening, and what I could do to prevent it?
I’m confused about what you’re doing and why you’re doing it. Are these EPI datasets that you are trying to register but showing them only with two colors, or is this image of a two-value dataset? That is what appears to be, and that’s a problem. The alignment uses a cost function that is based on a different intensity between CSF and other tissues and different between T1 and EPI data. The “ghosting” is probably more the result of interpolation. For continuous intensity datasets like EPI, these are typically very low intensities outside the head, so they are not very visible.
The align_epi_anat.py command might have some problems too, but it’s hard to say without knowing what is you are doing here. The anatomical dataset has a skull included, but you have said not to remove it with “-anat_has_skull no”. The other choices are also a little different. Do you want to save the motion correction here and the slice timing corrected datasets? If you are doing this as part of many typical FMRI processing streams, consider afni_proc.py to manage calling align_epi_anat.py and saving motion parameters and datasets.
This is a regular EPI dataset. I just chose this colour scheme in the overlay, because I find it easier to see if anatomical markers (e.g. ventricles) are lining up with the underlay. But this a regular EPI with continuous intensity. The anat I enter into align_epi_anat is also skull-stripped (and the skull-stripping is good). I’m doing volume registration, skull-stripping, etc in separate steps because align-epi-anat is crashing for a lot of my subjects when these steps are done together.
I’m just concerned about the voxels that fall outside the brain. These voxels are introduced in the alignment step because the epi before alignment looks okay. I’m also concerned about all the bleeding out in the normalised EPI. Is that normal?
Maybe a good reason not to use this kind of color scheme. I would guess the banding of the colors comes from the coil arrangement. Try looking at these with edges in a continuous color scale as shown in the afni10_volreg_talairach handout from our bootcamp classes. @AddEdge, the ‘e’ key over the viewer, @snapshot_volreg and a number of other tools can be used to asses alignment.
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.
