Hi Arash,
1. Add "-suffix NONE" to your @auto_tlrc command. If the orig
and tlrc anats do not have the same name, there might be no
assumed transformation between them (by afni).
You might be able to just rename the *_at anatomy, but I am
not positive. Try it.
2. The most simple way to transform results to tlrc space is
using either adwarp or @auto_tlrc, with the anat+tlrc dset
as the transformation parent. See the help examples, and
note that this is usage #2 in @auto_tlrc.
It's up to you whether to run 3dDeconvolve in orig space
or tlrc space. Maybe read over the "WARP TO TLRC NOTE"
section in "afni_proc.py -help".
3. I'm not sure why you wouldn't align all of your EPI data
together, and separately register the anats. Note that
3dvolreg should work for both of those steps, of course.
You could start by registering the anats, and then treat
everything like 4 runs plus an anat.
To register the anats, some do not like the idea of aligning
anat2 to anat1 and then averaging, since anat2 will be a bit
more blurred (since interpolated). To even this out, one
might extract the alignment matrix and apply the square root
to anat2 and the inverse square root to anat1, so they are
both transformed, meeting in the center say. Then average.
For "least interpolation", I don't see why you aren't already
there. By using -tlrc_apar in your align_epi_anat.py command,
the EPI data should already be in tlrc space (EPI alignment,
anatomical alignment and tlrc transformation all in one step,
and so one interpolation).
So it's odd that your subsequent 3dDespike command reads an
orig dataset. Does the +tlrc dset exist?
- rick