You could make the model more suitable for these data by
1) Only using one Gaussian per tissue class (except for the background class).
2) Including some noise, as the segmentation is not ideal for modelling PDFs
as delta functions.
Best regards,
-John
> i want to test spm for segmentation od sphere.
>
> i mead a sphere in difrent layres and each layer has its one gray level,
> for example 255 170 100 60 200 for 5 tissues in the head (WM, GM, CSF,
> Skull and skin).
>
> and i meade a templates for them. and i run the spm5b with this atlas.
> but spm could't segment the CSF correctly. spm segments CSF and skull as a
> one region.
>
> 1-
> why this happen?
>
> 2-
> if i use the skin or skull map instead of csf, for segmenting skull (in
> sphere example), i have two tissues in a one region, like: csf and skull
> or csf and skin!!!!
|