[caret-users] do spherical registration using freesurfer generated surfaces

Sandoval, Traci I tis031000 at utdallas.edu
Thu Sep 17 16:07:55 CDT 2009

Donna, I would love to use these scripts that aren't publicized yet. How can I attain said scripts?!! Thank you for all your work and questions answered. Traci

Traci Sandoval
Research Assistant
NeuroPsychometric Research Lab
School of Behavioral and Brain Sciences - University of Texas at Dallas 
Center for BrainHealth - University of Texas at Dallas
Department of Psychiatry - University of Texas Southwestern Medical

-----Original Message-----
From: caret-users-bounces at brainvis.wustl.edu [mailto:caret-users-bounces at brainvis.wustl.edu] On Behalf Of Donna Dierker
Sent: Thursday, September 17, 2009 8:44 AM
To: Caret, SureFit, and SuMS software users
Subject: Re: [caret-users] do spherical registration using freesurfer generated surfaces

Hi Jidan,

See inline replies below.


On 09/17/2009 02:06 AM, z丹丹 wrote:
> Hi,
> I already have fiducial surfaces generated by Freesurfer, which are
> all in the MNI space. I want to do spherical registration to an atlas
> spherical surface. The steps i need to do are:
> 1, generate the ellipsoid, spherical surfaces, and flat maps.
> 2, create the 6 landmarks on both individual and atlas surface.
> 3, do spherical mapping.
> Are the procedure right? Thanks! 
Funny you should ask. We (the Van Essen Lab) are using Freesurfer for no
less than five projects, and we have shell scripts that streamline much
of the import, border drawing, QA, and registration. There is a feature
we call "auto-landmarks" that draws the borders reasonably well. Many
borders need touch-ups, but a single rater fixed 72 hemispheres' borders
in 1.5 days. If you have only a handful of subjects, it's not clear
whether it would be worth it for you to use our methods. But if you have
dozens, then they would save you a lot of time.

We haven't broadly publicized these features/scripts, because they're
still in the final stages of development and testing. But they have been
working well for us.

Even if you don't use auto landmarks or these scripts, one question
stands out in your steps above: Do you need a flat map? No, not really.
It certainly is easier to draw landmarks on the flat map than the
sphere; however, I'm not sure it's worth the hassle. Then again, if
you're not using the scripts, it might be the shortest path to registration.
> In fact before I move to the 2nd step, I was stuck in the flat map
> generation. I met two problems:
> 1, As my surface is in MNI space, the surface orientation is not right
> in the CARET software. I found that the anterior and posterior are
> reversed with superior and inferior. Is there any method to roate it
> into the right orientation in CARET?
Yes, but this is not consistent with my experience of MNI space nor
Freesurfer surface orientation. While MNI volumes are sometimes stored
using left-handed coordinate systems (what some call radiological
orientation), the coordinates are always reported with -x being left, +x
being right; -y being posterior, +y being anterior; -z being inferior,
+z being superior. I've never seen a Freesurfer surface with the Y and Z
axes flipped -- not even ones written in MNI space.

It's easy to flip a surface from LAS to LPI
(http://brainmap.wustl.edu/OLD/SureFit/orient.html) like this:

caret_command -surface-apply-transformation-matrix $COORD $TOPO
$FLIPPED_COORD -matrix "1 0 0 0 0 -1 0 0 0 0 -1 0 0 0 0 1"

But I really would not brush this off. I would get to the bottom of why
the supposedly MNI surface is not in the expected orientation. The
evidence suggests it is not in MNI space, as asserted.
> 2, The other problem is , I found that the origin of my surface is not
> AC. So when I flatten the surface, it can't find the correct medial
> wall to cut. Do you think there is a way to solve this too?
This is more evidence that the surface is not really in MNI space. How
did it get transformed to MNI?

Note that the MNI origin is not exactly the AC; see
http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach for details. But
it is certainly close enough to the AC that the compressed medial wall
view should be centered over the medial wall. The fact that it is not
indicates a problem with your transformation to MNI space.
> Thanks a lot!
> Jidan

caret-users mailing list
caret-users at brainvis.wustl.edu

More information about the caret-users mailing list