> After segmentation process, I opened the segmented files in matlab (only
> the white and grey matter) and noticed some overlap between the two.
If you add up the values, then they should not exceed one at any voxel because
they represent probabilities of voxels belonging to any of the classes.
>
> Is there any criteria that I should use in order to decide to which matter
> the overlap area belongs to?
Not really. How these probabilities should be made into discrete values is a
little arbitrary. You could just assign each voxel to the most probable
class though. This could be by using ImCalc and an expression such as
(i1>i2) & (i1>i3) & (i1>(1-i1-i2-i3))
Best regards,
-John
|