What I have seen is that SPM5 segmentation is vulnerable to poor affine
registration of data and templates. It may help to reorient your images
roughly (+/- 10mm trans, +/- some degrees rot) into standard space using
CheckReg context menu item "Reorient" - you can display your image along
with one of SPMs templates and perform visually guided manual
coregistration. Alternatively, you can try to use SPMs coregister routine
to do this automatically.
Hope this helps,
Volkmar
On Mon, 9 Jul 2007, Ashburner John (PSYCHOLOGY) wrote:
> Is there any indication that there has been any sort of problem with how
> the tissue probability maps are overlayed on the images (ie check that
> the initial affine registration has worked). If there is a problem,
> then it should be extremely obvious - although sometimes people don't
> appear to notice. More accurate initial registration may be required
> (ie via the Display button).
>
> Other than that, I'm not really sure what the problem is with your data
> and why it doesn't seem to fit the SPM5 segmentation model so well.
>
> Best regards,
> -John
>
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Kenneth Rando
> Sent: Friday, July 06, 2007 7:41 PM
> To: [log in to unmask]
> Subject: [SPM] posterior probabilites in SPM5 segmentation
>
> Hello Listmembers,
>
> I have previously used SPM2 sucessfully, but I am using SPM5 for the
> first
> time to perform a VBM analysis. It is also the first time that I am
> using
> 3T scanner data. By comparison to our earlier 1.5T data the gray
> matter/white matter contrast is lower. Inhomogeneity has also been a
> problem with this data. Reducing the spatial frequency of the bias
> correction in the Segment function has helped, based on visual
> comparison
> of segmented data to the original scans.
>
> Compared to my SPM2/1.5T data results, I have noticed that the voxel
> posterior probabilites in the gray matter segments are lower. My SPM2
> gray
> matter segments contained mostly probabilities near 1 or 0 (which I
> expect,
> if I understand some of the published or SPM list discussions of
> segmentation). The cortical regions at the edges of the image, where
> the
> inhomogeneity has been a problem, have values more in the range of
> .75-.85
> (perhaps less). I can't fully correct the inhomogeneity problem, it
> seems,
> without inaccurately reclassifying gray matter as white matter. Does
> SPM5's Unified Segmentation contribute to the greater variability in the
>
> probability values? More importantly, can I still expect valid
> estimates
> of group effects?
>
> Thanks.
>
> Ken
>
--
Volkmar Glauche
-
Department of Neurology [log in to unmask]
Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
Breisacher Str. 64 Fax 49(0)761-270-5416
79106 Freiburg http://fbi.uniklinik-freiburg.de/
|