Hi Donna, John, Thank you very much for the previous suggestions. I managed to load+fix a manual segmentation, reconstruct the surface and register it to the f99ua1 surface.
I now deformed paint files from the atlas back to my volume (using "copy surface paint column to volume", and saved it as nifti). What I did not find out is which paint labels the integer numbers in the paint volume refer to. Is there a file I can look this up? Or is there a way to save this information in caret? Thanks, -- Dirk On 4/20/07, Donna Dierker <[EMAIL PROTECTED]> wrote:
Hi Dirk, See inline replies below. Donna On 04/20/2007 12:59 AM, Dirk Neumann wrote: > Dear All, > > We got a high-resolution (.15 mm) T2-weighted scan of a macaque brain > (gray matter brighter, white matter darker) and would like to > reconstruct the cortical surfaces and align it to a standard atlas. > > Could this be done with any available software? I'm sure, but the ones I know of haven't been released for public consumption. > > I had a look at the caret/surefit source code. Would you recommend > adapting the generateInnerBoundary / generateOuterBoundary / > generateSegmentation functions ? Let's put it this way: Even back in the days when SureFit was in python, and it was pretty easy to just try different things, I never did this in those "all I have is T2" (or my T1 is crummy) situations. Instead, I'd use the equivalent of Caret's Window: Script Builder functions to try to transform the T2 into something that looks like a T1. This usually involved somehow getting masks of the intracranial/extracranial voxels; reversing intensities; and dilating/eroding to get rid of white residual voxels along the edges. Not pretty, not ideal, not fun. > > Would there be an easy way of providing caret/surefit with a manually > segmented volume or reconstructed surfaces (from amira/itk-snap) ? Sure. You can do this. Use Volume: Open Anatomy Volume and open your structural; then use Volume: Open Segmentation Volume and open your amira/itk-snap output. Use D/C Overlay/Underlay volume and set anatomy as underlay, segmentation as overlay; make sure they align. If not, adjust orientation/origin as needed to make them align. If you just want a surface, you can do Volume: Segmentation: Reconstruct into surface. But if you want to carry it through to registration, then you'll probably need to first make sure: * The segmentation contains a single, binary object with intensities 0 or 255, no handles (or only tiny ones). Use Volume: Segmentation: Topology error report to check number of objects, cavities, and handles. * Both volume and segmentation are centered somewhere near the AC. (Later, after you've generated your surface and registered to an atlas, you can create an alternate version of your fiducial surface that is centered on an alternate origin, e.g., interaural.) Otherwise, when you flatten the compressed medial wall will be centered in an awkward location. You could, however, use the very inflated surface to draw your borders instead. Anyway, if you're registering the surface, use Volume : SureFit Operations: and make sure volume type is segmentation. Then turn off all options through auto error correction, but turn on Generate fiducial surface and everything below it. > > Thanks, -- Dirk > > ____ > > Dirk Neumann > Computation and Neural Systems Program > California Institute of Technology > (626) 395 4025 dirk @ caltech . edu > http://emotion.caltech.edu/~dirk/ > _______________________________________________ > caret-users mailing list > caret-users@brainvis.wustl.edu > http://pulvinar.wustl.edu/mailman/listinfo/caret-users _______________________________________________ caret-users mailing list caret-users@brainvis.wustl.edu http://pulvinar.wustl.edu/mailman/listinfo/caret-users