The initial affine registration is not as robust as it would be if there were
a few non-brain tissue probability maps in existance (bone, scalp etc). This
lack of robustness means that the data need to be in reasonable alignment
before running the code. Alternatively, skull-stripped data could be used
(as these do not need such tissue probability maps in order to have a good
model of the image intensities).
Best regards,
-John
On Friday 12 October 2007 18:03, Paul Macey wrote:
> To follow from Ged, I use the attached script to calculate the x-y-z
> shift to apply to the image so that it overlaps the MNI template (see
> help/comments on applying to the image).
>
> Best wishes,
> Paul
>
> DRC SPM wrote:
> > Hi Amir,
> >
> > This is probably an issue that has come up on the mailing list a
> > couple of time, briefly: SPM5's unified segmentation is more sensitive
> > to the initial rigid alignment of the images with the template/priors
> > than SPM2's normalisation was.
> >
> > Use Check Reg, with the template and your images, right click your
> > image and select "reorient this image" from the context menu (I'm
> > working from memory here, but hopefully that's correct). You don't
> > need perfect AC-PC alignment, but you do need rough agreement of brain
> > orientation.
> >
> > Hope that helps,
> > Ged.
> >
> > On 12/10/2007, Amir Tahmasebi <[log in to unmask]> wrote:
> >> Dear SPM experts,
> >>
> >> I am using SPM5's unified segmentation technique for warping. It works
> >> fine on 17 out of 21 subjects but it doesn't properly deform the other 4
> >> subjects' data. However, when using SPM2's normalization, all 21
> >> subjects are properly deformed to match ICBM template. I tried different
> >> values (1, 10, 100) for the warping regularisation but still couldn't
> >> make it work. I appreciate it if you could kindly help me with this
> >> case. I can send you one of the datasets if you want to try. Thanks in
> >> advance.
> >>
> >> Regards,
> >> Amir Tahmasebi
|