Question about reference image volume for alignment

AFNI Message Board
1

Dear AFNI experts,

The default script generated by afni_proc.py was using the second volume of functional image
of the first functional run in align_epi_anat.py with options, -epi pb01.$subj.r01.tshift+orig -epi_base 2
Also, in registering volumes, in 3dvolreg, with option, -base pb01.$subj.r01.tshift+orig’[2]’

I have 4 functions runs with short break between runs using the second volume of the first run (i.e., r01),
as a reference volume for all 4 runs for alignment and volume registration.

However, I wonder if using the second volume for EACH of run separately is better idea than the default script is doing.
Is there any difference between the two alignment methods?
(using one reference volume from the first run vs. using reference volume for each of run separately).

I already finished preprocessing using the default script, and wondering if I should do it again separately for each run.

Sincerely,
Sungshin Kim

2

Hi Sungshin Kim,

Even if you align each run to a separate base, the
runs still need to be aligned together some way.
Using EPI volumes for that step, it is not clear why
that would yeild a very different result. Using the
anatomical volume for it, perhaps it would be hard
to predict. With great data, it might work fine.
Otherwise, it could easily be worse.

What might be beneficial is using the MIN_OUTLIER
volume, which I recommend. That is pretty sure to
be a low-motion volume, which should lead to more
robust registration.

  • rick
3

Thank you for your answer, but I need to make something for sure.

Based on your answer, it seems that between the two options, using one reference from the first run for all the images would be better than using
separate reference for each run…right?
Also, these two approaches would make a significant difference, but we don’t clearly understand the reason, right?

Thanks.

4

It is hard to answer that because you have not yet
said how you plan to get alignment across runs. Is
it by aligning each run to the anat?

It would be more accurate to say we could not
predict the difference in general than to say we do
not understand why there might be a difference.
This is a data-dependent issue.

  • rick