Dear afni experts,
I have a question about the MapIcosahedron function and in general how to link original surfaces and std surfaces.
I understand that MapIcosahedron looks for the registered surfaces (e.g., lh.sphere.reg) and re-mesh this spherical surfaces with a standard number of nodes. Different subjects might have different numbers of nodes in their lh.sphere.reg but once you re-mesh the surfaces different subjects should have the same number of nodes in their std.lh.sphere.reg.gii. However, this is just for the sphere surface, how you further create other surfaces, like, std.lh.pial.gii, std.lh.face.gii etc? In freesurfer, we typically first obtain the surface like lh.white and then inflate it in to lh.sphere. But now you first obtain the sphere surface, how you can fold it into std.lh.white.gii??
I assume you might use the nodes mapping file like std.lh.MI.1D though I do not understand how you exact use it. I found in this file each row has 7 columns. The first col is the node number in the std surfaces and the col 2-4 are the nodes in the original surfaces. What are the cols 5-7?
Thanks so much.
How this is done is lightly described in the suma.pdf class handout, starting on slide 30 (with “our way” on slide 32). The basic point is that the icosahedron is mapped to the registered sphere, making a mapping of each icosahedron node to a location in the brain (one location on the sphere is supposedly a consistent anatomical location across subjects). So each iso node is a “fixed” location in the brain across subjects.
The next step is to regenerate all of the original surfaces using the mesh of the fixed icosahedron. Since the subject’s sphere gets warped to align with the standard one, it’s corresponding smoothwm, pial or whatever surface nodes have mappings to known locations on that sphere. So for each node on the ico, its location on the warped sphere will pierce a triangle of one of the original surfaces (e.g. smoothwm). From that location, that node will get the interpolated coordinates from the nodes of the pierced triangle, based on proximity to each of the triangle’s vertices.
That creates a new surface, with the mesh of the icosahedron (our “standard mesh”), but the coordinates of the original space surface (e.g. smoothwm). It will look almost identical to the original space surface. And it is now ready for group analysis.
The original surfaces are then promptly ignored, and all AFNI/SUMA processing is applied with the standard mesh surfaces. One can even display the group results on a corresponding standard mesh surface from the MNI152 anatomy, and now the coordinates are in MNI space, but still on the same standard mesh.
Does that seem reasonable?
Thanks Rick! This is super helpful!