On Nov 13, 2006, at 11:26 AM, John Arsenault wrote:
Firstly I found that I got a better segmentation with the .35mm
isotropic voxels relative to the same volume resampled to .5mm
Good to know.
Secondly I am having a bit of trouble registering an
individual hemisphere to any of the new 73700 node-standard-mesh
representations of the three atlases. I am using the
“Macaque.F99UA1.REGISTER-with-INDIVIDUAL.73730.spec” to spherically
deform to. I have had trouble get any of the .borderproj from this
.spec file to deform to similiar (same type, number, and
landmarks) .borderproj files I create with my indivual hemisphere.
Hmmm. Those are certainly the things to check first.
One thing you might possibly have overlooked is if there is a stray
border point that inadvertently got projected to the origin somewhere
along the line (e.g., because it sits along a cut in the flat map).
The way to check this is to view the borders on any 3D surface (e.g.,
the sphere) and to choose the Page Selection: Borders: Draw Borders
as Points and Lines pulldown option. Then check to see that none of
the borders has any irregular trajectories.
am not sure at which level my mistakes are happening but I am getting
very high numbers of crossovers which stop me from completing
deformation. When I have the ATLAS and the individual open in
windows I have the same number(also confirmed by .spec file), type and
orientation(denoted by red point) in the individual as with the atlas.
I then go to deform spherical in the individual and choose the
apropriate files but I am a bit confused on which fiducial and which
.borderproj to pick for the atlas:
Fiducial Coord(choose one of following):
Macaque.F99UA1.RIGHT.FIDUCIAL.Std-MESH.73730.coord(I use this one)
This should NOT matter at all IF (as I recommend) you DE-select the
'Correct for spherical distortion relative to fiducial' in the
Spherical Parameters window. This correction option is one that we
implemented years ago, but I was never able to convince myself that
it actually performed as intended.
Even if you do select it, I don't think it should make much
difference or be the cause of your problems.
.borderproj(I normall use a modified version of Surfs with only 14
Not sure what this means - are you using only 14 landmarks instead of
the 23 - 26 in the set below?
These differ from one another in modest ways because each was
customized to reflect the landmarks that could be reliably identified
in the source and target pairs.
You should choose whichever set of landmarks best matches the
individual you're trying to register, then create an atlas spec file
containing ONLY that borderproj file. OR if you need to modify the
landmarks in the atlas to match what you've settled on for your
individual, then create a new atlas borderproj and a spec file with
only this new set.
As well I know the new landmarks are averages of left and right
but can I use the right flat or right fiducial to deform to left
You should be able to register between either a left or a right hem
to the atlas 'LR' target. But make sure that your spec files and the
coord files therein all have the correct hemisphere assignment -
otherwise troubles can ensue.
Not sure if any of this is causing all of the
crossovers I am experiencing.
Well, you've now got a few leads to follow. If none of them pans
out, you can ask more questions. Or at some point it might be most
efficient to upload the files and let one of us look for the
problem. But I don't think we're to that stage yet.
Thank you for your help,
caret-users mailing list