this seems to be a problem with intensity inhomogeneity
which seems to be far more intense than SPM method assumes.
you can try to correct this bias field prior the segmentation with
I have developed a method to do so. you can find it at:
it is a bit slow but it is very general and robust.
hope this helps
Mensaje citado por David Ruhl <[log in to unmask]>:
> I have a problem segmenting a subset of my subjects' images. For most
> subjects everything goes smoothly, for some (~20%) the segmented images come
> out REALLY strange (see attached - one subject's T1 and three segmented
> images). I've rerun it several times, screwed with the parameters, and am
> out of ideas. Anyone know how to fix this?
> What was done to the images in spm:
> Dicom import of T1 & T2
> Coregistration of T1 to T2 (axial)
> Segmentation of gm, wm, & csf into native space (This is where the problem
> David Ruhl