Question Regarding AFNI Proc Registration Options

AFNI Message Board
1

I employed the options listed below when processing intra-session data from a single subject.
-volreg_align_to third
-volreg_align_e2a
-volreg_allin_cost lpa+zz
-volreg_post_vr_allin yes
-volreg_pvra_base_index MIN_OUTLIER
-align_opts_aea -AddEdge
-align_opts_area -giant_move

My take on the tcsh script is that it first computes the anat alignment transformation to EPI registration base. Then it does the motion correction for each run and then cross-run alignment of reg bases (uses the min outlier for each run and it seems to use an 0 as the overall base). After this it does the epi2anat registration. Is my understanding correct? Also, I specified align to third for the motion correction but it is fine if the overall base is 0 I assume the other options overrode that specification.

2

Hello,

I do not think it used base index 0, except that it had already extracted index 3 into a new, single-volume vr_base dataset, from which it then did use index 0. So ‘third’ should still be having an effect.

On that note, if you are going to use MIN_OUTLIER across runs, it would make sense to also use ‘-volreg_align_to MIN_OUTLIER’.

But otherwise, yes, your understanding seems correct. Since it already knows how to align the EPI and anat, the transformations to the EPI volumes (based on the given options) become:

current EPI volume → per run base → overall EPI base → anat

  • rick
3

Thanks for making that clear. I was thinking about changing the base 3 to min outlier as well. Although I am not sure if it will make a major difference.

4

It might or might not make a difference. The important case comes when a subject moves during time point #3, which might mess up that volume. That is why MIN_OUTLIER is used.

  • rick
5

Makes sense I will try it again using Min outlier for both. The t maps from the model look good. I have tents and block predictors. The only strange thing is for a get a lot outside the brain, which is normal, but even at really low q values (.001) for some t maps. One of my block predictors has some deactivation outside the brain and some strong OFC activation despite the weak TSNR at that location at q=.001. The activation from the tents seems to show up were I would expect.