 |
 |
> I am getting into trouble with optimized VBM and T1 images of elderly > subjects. When hi-resoution (1.7 x 0.82 x 0.82mm voxel) images were > segmented, gray and csf are not properly segmented. Seg1 image appears to > include both grey and csf, and seg3 image has only outer brain tisssues. > Please see an attached jpeg image. Among 11 brains processed, the miss- > segmentation occured in four images. These four T1 images do not look very > difefrent from other properly segmented images. It looks like something is going wrong with the spatial normalisation part of the procedure (or the initial affine registration). The tissue probability maps (priors) are not being properly overlayed, so the segmentation does not work so well because the wrong priors are used. It looks like the whole head of the subject has been matched to the brain of the template image. In SPM2, the spatial normalisation (or the initial affine registration of the Segmentation) matches the image to a template of the same modality. This can go wrong - especially if the contrast in the templates differs from that of the subjects images, or in cases where the subject has a thinner skull than those in the images that constitute the template. You may find the procedures within SPM5b work better for VBM preprocessing - although they are not perfect. I have a dataset at the moment that I am really struggling to segment properly. Best regards, -John