The segmentation sometimes needs very good starting estimates (via
image reorienting) before it gives a good solution. This is because
the initial affine registration is sometimes not so robust because it
does not use any information about non-brain signal in the image. If
you check the image segmentation results (eg c1*.img, c2*.img), I
would expect that these may not have worked so well.
The new segment is generally much more robust, but I have not yet got
around to making it easier to use for spatially normalising images.
Best regards,
-John
On 7 December 2010 19:00, Asaf Kaftory <[log in to unmask]> wrote:
> Hello dear SPM'ers,
> we've encountered a strange problem in the segmentation phase in our pre-processing (done with SPM8):
> When the normalizing is done using the parameters from the segmentation (i.e., parameter file is the file seg_sn.mat) we get an extremely distorted image. However, if the normalization is done using the T1 image after segmentation (i.e., the anatomical image with the prefix m), we get a good image.
>
> would love to hear your thoughts on the subject...
>
> thanks,
> Asaf.
>
|