| I have some T1 images I have normalised. The normalised images look
| sensible with good registration. However, when I segment the images,
| I appear to get an underestimation of grey matter and an excess of CSF.
| This particularly affects medial frontal regions and the anterior
| cingulate, where in comparison to the normalised images, the
| interhemispheric fissure seems far larger in the segmented images.
| Is it possible to "tweak" the segmentation?
What resolution are you spatially normalising your images to? If it
is to only about 2mm resolution, you could try using smaller voxel
size in order to reduce partial volume effects. The classification
model does not handle partial volume so well, as voxels are assumed
to belong to only one class. When voxels contain a mixture of different
tissues, they may be more likely to be assigned to the class with the
highest intensity variance - in this case, CSF rather than grey matter.
Best regards,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|