I confess I don't follow this 100%, but if you are faced with monkey segmentation challenges, Colin's recipe might help.
Begin forwarded message: > From: Colin Reveley <[email protected]> > Date: November 20, 2011 11:20:16 AM CST > Subject: homoeomorphic to sphere > > > the problem of a minor topological defect (so minor I still don't know where > it is) resulting in freesurfer filling in the surfaces to a solid block for > the entire occipital lobe (which obviously won't do) has a simple solution. > > freesurfer is useful in making WM and pial surfaces, and also "freeview" is > really nice for looking at surfaces as diagrams in each plane for comparison > to e.g. paxinos atlas, and targeting roi's as volumes for e.g. seeded > tractography with ref to a paper. freesurfer is also more manual giving more > freedom if you know what you are doing. Which I don't. but I did get a nice > WM and pial line. > > If you don't want those things, don't bother probably. Although, the > midthickness generated in caret from FS white and pial is more to my liking > than surefit (it looks more like F99) , but that is probably because of my > data properties (ex-vivo MTR, rather than in-vivo T1). > > To fix: > > adjust once the 3 surfaces (white, pial, mid) are converted/produced and > imported to caret (using commands in e.g. recent VE FS-F99 tutorial). Then, > you just can use the GUI and select "crossover smoothing" in the surface > menu. And they are fixed. I still cannot see the difference before and after > to be honest; it's pretty subtle. > > you could, I'm sure, just import the FS "orig" surface (the first go at a > white matter surface), fix it, convert back to freesurfer format and > circumvent the entire awfulness of FS spherical topology fixing. I spent a > long time with that (inflating it more, less etc). no dice. > > in caret, smooth crossovers on the surfaces works wonderfully, only a few > thousand edges changed of more than 200,000. > > This has been work. I'm now pleased with what I have, perfectionist that I > am. Do contact me, and I'll advise with what ad-hoc knowledge I have. > > generally: avoid mris_sphere -q, mris_fix_topology process for macaque if it > doesn't work off the bat. use CARET crossover smoothing. > > Colin, Sackler Centre, Sussex
_______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
