and where do you get ground truth? On Wed, 30 Dec 2009, Zhong Jidan wrote:

Hi, I want to check the deformation accuracy of FreeSurfer in this way.. By calculating the deformation vectors' difference before and after mapping, I can know which part is registered better with the reference of the true ground.

Thanks,

Jidan

On Tue, Dec 29, 2009 at 11:01 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:

...

Hi Jidan,

can you give us an idea of what your goal is. Lilla (ccd) does some of what you want as part of her CVS processing stream.

cheers, Bruce

On Tue, 29 Dec 2009, Zhong Jidan wrote:

Hi Freesurfer experts,

...

I have a 3D fiducial surface ('anatomical' surface configuration), based on this fiducial surface ,let's say, "template", I generated some simulated fiducial surface with some deformation, let's say "subject". So here I know the "deformation vectors" for each vertex between the template and the subject in the original 3D fiducial surface "space". Then I use FreeSurfer to generate the spheres and do spherical registration from those subjects to the template. I want to check how these " deformation vectors " I got from Freesurfer are different from my generated "deformation vectors". But the problem for me is, after Freesurfer registration, the " deformation vectors" is based on the spherical space. While my generated "deformation vectors" are in the 3d fiducial surface space, it's not comparible for these two vectors. I want to know whether I can put the deformed subject sphere back into the fiducial surface representation. In that way, I can calculate the difference in the fiducial space. If getting the deformed subject sphere into the fiducial surface space is not possible, could you give me some suggestions about how to calculate the difference of the deformation vectors which are not in the same space? Generating the deformation vectors in the spherical space is the last thing I want to do because you never know how it likes in the real fiducial surface representation.

Thank you.

Jidan

On Mon, May 4, 2009 at 8:48 PM, Zhong Jidan jidanz@gmail.com wrote:

Hi,

...

I have a question about the sphere registration. The registration of surface is spherical registation, so the outpput of the surface is also a sphere. When we have the inflated and smoothwm, we can convert the surface into sphere using mris_sphere, but reversely, how to convert the deformed sphere in the form of white surface or pial surface?

Also, if I want to use the surface between the white and pial surface to do registration, do I need to make the sphere myself?

Thanks,

-- Regards,

Jidan

I see, and you want the spatial offset vector associated with each vertex on the surface in anatomical (as opposed to spherical) coordinates? Can't you generate a spherical parameterization for both your subject and your template, then use some type of interpolation (e.g. barycentric) to compute the weighted average of the vertex coordinates in the template face that each source vertex maps to? Lilla probably has some code like this around.

cheers, Bruce

On Sat, 2 Jan 2010, Zhong Jidan wrote:

Hi,

My 'subject' surface, which are simulated fiducial surface generated with some deformation based on the template, has the one-to-one correspondance with the template surface. This can be seen as the ground truth.

thanks

jidan

On Thu, Dec 31, 2009 at 11:38 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:

...

and where do you get ground truth?

On Wed, 30 Dec 2009, Zhong Jidan wrote:

Hi,

...

I want to check the deformation accuracy of FreeSurfer in this way.. By calculating the deformation vectors' difference before and after mapping, I can know which part is registered better with the reference of the true ground.

Thanks,

Jidan

On Tue, Dec 29, 2009 at 11:01 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:

Hi Jidan,

...

can you give us an idea of what your goal is. Lilla (ccd) does some of what you want as part of her CVS processing stream.

cheers, Bruce

On Tue, 29 Dec 2009, Zhong Jidan wrote:

Hi Freesurfer experts,

...

I have a 3D fiducial surface ('anatomical' surface configuration), based on this fiducial surface ,let's say, "template", I generated some simulated fiducial surface with some deformation, let's say "subject". So here I know the "deformation vectors" for each vertex between the template and the subject in the original 3D fiducial surface "space". Then I use FreeSurfer to generate the spheres and do spherical registration from those subjects to the template. I want to check how these " deformation vectors " I got from Freesurfer are different from my generated "deformation vectors". But the problem for me is, after Freesurfer registration, the " deformation vectors" is based on the spherical space. While my generated "deformation vectors" are in the 3d fiducial surface space, it's not comparible for these two vectors. I want to know whether I can put the deformed subject sphere back into the fiducial surface representation. In that way, I can calculate the difference in the fiducial space. If getting the deformed subject sphere into the fiducial surface space is not possible, could you give me some suggestions about how to calculate the difference of the deformation vectors which are not in the same space? Generating the deformation vectors in the spherical space is the last thing I want to do because you never know how it likes in the real fiducial surface representation.

Thank you.

Jidan

On Mon, May 4, 2009 at 8:48 PM, Zhong Jidan jidanz@gmail.com wrote:

Hi,

...

I have a question about the sphere registration. The registration of surface is spherical registation, so the outpput of the surface is also a sphere. When we have the inflated and smoothwm, we can convert the surface into sphere using mris_sphere, but reversely, how to convert the deformed sphere in the form of white surface or pial surface?

Also, if I want to use the surface between the white and pial surface to do registration, do I need to make the sphere myself?

Thanks,

-- Regards,

Jidan