Hammers, Alexander wrote:
> I believe that the reason is the partial volume effect (yes, not exclusive
> to PET /SPECT J ) – low-intensity ventricular CSF next to high-intensity
> WM will lead to a rim of mid-intensity voxels for the usual 1-2mm linear
> voxel sizes, and will have intensity similar to the GM position in the
> histogram. This idea is also in keeping with the finding that tissue
> segmentations other than SPM lead to the same misclassification.
Seems very likely to me. I also see SPM-/FAST- segmented GM around the
brainstem, e.g. quite noticeable anterior of the pons -- am I right in
thinking that this is also anatomically wrong? If so, the same
explanation of partial volume effect on the WM/CSF boundary would make
sense.
Best,
Ged.
|