I'm not sure which version of SPM you are using, but there can be problems
with the registration part of the segmentation if your image is corrupted by
a lot of intensity nonuniformity.
I would suggest trying it with SPM5. Hopefully it is much simpler. The new
set of updates should (I hope) make the procedure more robust and stable.
Best regards,
-John
> I had a T1 image in native space and in order to apply an optimal
> transformation parameter set to normalize my EPI images into standard
> space, I tried to segment my T1 image in native space first and then
> normalize the segmented T1 image to the grey.mnc. Unfortunately, the
> segmentation failed. It looked like the segmented grey matter is much
> smaller than it was supposed to be, especially in the visual cortex field.
> If I check
> registration with the segmented grey matter image and orignial T1 image, it
> can be clearly seen that the visual cortex is not fully covered in the grey
> matter image.
> In this case, when later I tried to normalize individual T1 to template
> T1 to bypass the segmentation problem, the normalization quality was also
> pretty bad. Just wondering whether there are any hints to set up a
> successful segmentation? I was thinking about using the optimized VBM
> protocal: after I have the segmented grey matter image in native space, I
> can apply the transformation paramter set to individual T1 image to
> transform it into standard space. Then I can use segmentation again to get
> the grey matter image out of standardrized T1 image, which can be followed
> by the application of reversed ***_3dsn.mat to convert the better
> segemented image into native space. From there, maybe I can generate new
> transformation parameter set by normalize the "back-transfered" grey matter
> image to the template grey.mnc image.
> Is that a possible approach? I would really appreciate if somebody can
> offer any adivces.
|