afni proc, align_opts_aea not correctly applied

AFNI Message Board
1

Hey afni folks,

I am attempting to use 3dSkullStrip rather than 3dAutomask for anatomical to epi alignment, via


-align_opts_aea -cost lpc+ZZ -epi_strip 3dSkullStrip

When I check the proc script produced, the align section shows the align_epi_anat.py call contains both -epi_strip options, the default (first) and my request - but looking in the script output file, only 3dAutoMask is used to produce the ‘skull stripped’ epi,


3dAutoMask -apply_prefix ./__tt_vr_base_ts_rs_ns ./__tt_vr_base_ts_rs+orig

I can imagine a work around (skull strip the epi, and use that as a the volreg target) but I wanted to bring this to your attention. Perhaps I have done something wrong in the way that it is called, by placing the -epi_strip call last?

Thanks!

2

If only I had scrolled down slightly in the afni proc help, rather than jumping to align_epi_anat.

Solution is already built in, via


-align_epi_strip_method METHOD : specify EPI skull strip method in AEA

The AFNI wizards continue to think of (almost?) everything…

3

The automask method is the default in afni_proc.py for masking the EPI dataset. Have you run into a situation where it doesn’t work? Skullstripping for the anatomical dataset is done typically with 3dSkullstrip, or the newer @SSwarper.

4

I wanted to use 3dskullstrip on the epi data as part of align_epi_anat, as I thought 3dautomask was causing problems and leading to poor registration. In the end, turning off deoblique (-deoblique off) seemed to resolve the problem, strangely enough.

It was a troublesome and strange problem, as the whole brain epi and anat were very closely aligned to start with, nearly matched, but in some cases align_epi_anat was producing extraordinarily bad alignments, shifting half the brain out of the FOV, and stretching wildly.

@SSwarper worked well for 58/60 of the subjects - on two it appears to have ran into an issue during skull stripping - leaving far too much tissue around the brain, such that the MNI warped brain is very small. I’m still looking into that.

5

I am having the same problem with @SSwaper not fully stripping the skull in some datasets, leading to skull being warped as “brain.” Were you able to find a solution to this? I don’t see any options that would deal with this in 3dSkullStrip to pass to @SSwarper via the -ssopt.

Thanks!
Becca

6

Hi, Becca-

There have been some recent updates to @SSwarper that might address this; what is your AFNI version, i.e., the output of


afni -ver

?

AFNI versions of 20.0.03 and higher should have this update. (Though I note there are a couple further things I have been thinking about adding, too, but am Bootcamping at the moment so it will be a little while before I get to those.)

Running


@update.afni.binaries -d

–pt

7

afni -ver
Precompiled binary linux_openmp_64: Mar 15 2019 (Version AFNI_19.0.25 ‘Tiberius’)

I will try updating the binaries and see if that helps.

Thanks!
Becca