Creating the sphere may not be the first step, but the key thing about an atlas is that all of the subjects used to generate the atlas need to be on the same mesh, with subject landmarks occupying the same or nearby node numbers. Native meshes from freesurfer do not have this property, nor to surfaces generated from segmentation volumes. Those kinds of surfaces have the additional caveat that their node spacing and number of neighbors per node can be highly variable.
My understanding of how we make atlases (I haven't been involved in it, but it has been described to me) is that you draw landmarks on all of the subjects that make the atlas, unproject them to be borders that aren't tied to any surface, then resample and average the subject borders together (flipping left to be right if you want an atlas that has correspondence between right and left), and finally project those borders to the sphere that is the mesh you will use (again, flip the borders and project to the left sphere to get your left atlas target). Then, you can register your atlas subjects to this sphere, giving their surfaces node corrrespondence, and average their coordinate files to get the atlas average fiducial. Take this with a large grain of salt, as I said, I haven't actually done this. Tim On Fri, Feb 17, 2012 at 12:09 AM, Tristan Chaplin <[email protected] > wrote: > Thanks for information, but I must confess I don't understand why you > create the sphere first. I thought the procedure for atlases was to make a > surface, then resample it as a standard mesh, then do spherical morphing > etc. Is the idea instead to create the fiducial surface, do spherical > morphing, then align the sphere to one of these standard spheres? > > FYI we've already made a fiducial surface with cytoarchitecure as paint. > > > On Fri, Feb 17, 2012 at 16:08, Timothy Coalson <[email protected]> wrote: > >> We have moved away from the 73730 mesh, we are now using a new method to >> generate meshes which results in much more regular node spacing. Making a >> sphere is actually relatively easy, especially with the new release of >> caret. The hard part is making it into an atlas, which I defer to someone >> else. The command: >> >> caret_command -surface-create-spheres >> >> Will generate a pair of matched left/right spheres (mirror node >> correspondence, topologies with normals oriented out). I think that >> command made it into the 5.65 release, if not you can use spec file change >> resolution, and grab just the new sphere, and ditch the rest. The odd bit >> about spec file change resolution, though, is if you give it an old node >> count, like 73730, it will give you the old sphere (this is in case someone >> is relying on its old behavior). However, ask it for 73731 nodes, and you >> will get a new highly regular sphere instead (though it won't have 73730 >> nodes, because the 73730 node mesh wasn't a regularly divided geodesic >> sphere, but it will give you something close). If all else fails, there >> are a few spheres in the caret data directory. >> >> Tim >> >> >> On Thu, Feb 16, 2012 at 6:56 PM, Tristan Chaplin < >> [email protected]> wrote: >> >>> Hi, >>> >>> A while back I asked about creating standard mesh of 73,730 nodes, >>> similar to what is used for PALS atlas. I never got a chance to follow it >>> up then but I'd like to give it a go now. It seemed at the time that the >>> knowledge for creating such meshes was limited to a select few so if anyone >>> has any experience with this or has the contact details of someone I would >>> greatly appreciate hearing from them. >>> >>> The reason for creating this mesh is for making atlas for the marmoset >>> monkey. We are very interested registering this atlas to the macaque >>> monkey and doing analyses similar to Hill et al. (2010). >>> >>> Thanks, >>> Tristan Chaplin >>> >>> On Mon, Feb 7, 2011 at 16:04, Tristan Chaplin <[email protected] >>> > wrote: >>> >>>> Ok thanks for the information. >>>> >>>> >>>> On Fri, Feb 4, 2011 at 03:25, Donna Dierker >>>> <[email protected]>wrote: >>>> >>>>> On 02/01/2011 07:31 PM, Tristan Chaplin wrote: >>>>> > Hi, >>>>> > >>>>> > I've been reading about the creation of your atlases, and I see that >>>>> > PALS and the macaque atlases have standard size mesh of 73,730 nodes. >>>>> > I was wondering, is this the same across species to allow >>>>> > interspecies registration? i.e. is it still possible to do >>>>> > interspecies comparisons of other species with different size meshes? >>>>> Possible, but more difficult. Not to say that achieving vertex >>>>> correspondence across species is trivial. Interspecies comparisons are >>>>> really hard. I think David Van Essen is the only one in our lab that >>>>> is >>>>> doing them, although Matt Glasser might also be doing some. >>>>> > >>>>> > I was also wondering how the standard mesh was was actually made. >>>>> The >>>>> > PALS paper refers to the Saad 2004 paper, which I think uses SUMA. >>>>> > SUMA has a program called MapIcosahedron to create standard meshes. >>>>> > Is this still how you would recommend making a standard mesh? >>>>> Tim Coalson (a student who works summers here) also developed a utility >>>>> that creates meshes of specified resolution. >>>>> >>>>> Making a standard mesh is not something I ever do. You do it with a >>>>> specific motivation -- typically some other important data is already >>>>> available on that mesh. And the way you usually get your data on that >>>>> mesh is to register it to an atlas target already on that mesh. >>>>> >>>>> If you are talking about creating, say, a sparser mesh for mice/rats, >>>>> then you're out of my orbit. >>>>> > >>>>> > Thanks, >>>>> > Tristan >>>>> > >>>>> ------------------------------------------------------------------------ >>>>> > >>>>> > _______________________________________________ >>>>> > caret-users mailing list >>>>> > [email protected] >>>>> > http://brainvis.wustl.edu/mailman/listinfo/caret-users >>>>> > >>>>> >>>>> _______________________________________________ >>>>> caret-users mailing list >>>>> [email protected] >>>>> http://brainvis.wustl.edu/mailman/listinfo/caret-users >>>>> >>>> >>>> >>> >>> _______________________________________________ >>> caret-users mailing list >>> [email protected] >>> http://brainvis.wustl.edu/mailman/listinfo/caret-users >>> >>> >> >> _______________________________________________ >> caret-users mailing list >> [email protected] >> http://brainvis.wustl.edu/mailman/listinfo/caret-users >> >> > > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users > >
_______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
