Hi Ajay,

Here is a lot of babbling, but at the end there
is a conclusion...


The @SUMA_Align_To_Experiment command run by the
afni_proc.py script will only use an affine
transformation to bring a surface volume into
alignment with the current anatomy. There is no
current mechanism for applying a subsequent
transformation to the surfaces to take them to a
standard space, particularly using a non-linear
transformation.

That means as it stands, your single subject
surfaces would not be in alignment with their
(non-linear) standard space anatomies. So
they would not be exactly comparable.


A surface analysis is generally run in original
coordinate space. It is the surface topology
after @SUMA_Make_Spec_FS automatically runs
MapIcosahedron that puts the surfaces on a
"standard mesh". It is not the coordinates
that are standard, it is the node indices and
mesh (according to FreeSurfer's surfaced-based
registration to their own spherical template).

This means that a surface analysis run in orig
space with standard meshes can be used in a
group analysis with no extra step. However to
get the node coordinates into a standard space
(usually an unnecessary step), the group results
should be viewed on a surface that is actually
from the template space.

There are 2 ways to do that.

1. Make standard mesh surfaces from the anatomical
template, and report surface group results from
those surfaces.

This is usually sufficient, but since you care
about both volumetric and surface-based analyses,
keep in mind that the surface-based standard space
transformations would not match the volumetric ones.
One registration is done by FreeSurfer on the
spherical surfaces, one is done by AFNI on the
anatomical volumes. So the group results at any
given coordinate would not exactly match, though
they should presumably be similar.


2. Make single subject surfaces from the standard
space subject anatomies.

This way, not just the surface nodes but their
actual coordinates are in standard space. This
way, they volumetric and surface-based results
should more closely match. They still won't be
identical (blurring and interpolation happen
differently in the 2 domains), but they should
be closer than when using the previous method.

- rick



Edited 1 time(s). Last edit at 11/23/2015 11:23AM by rick reynolds.