Hi Pia.
> Hope you well and happy.
Can't grumble - but it never stops me.
> I like SPM8, well done.
Thanks.
>
> with respect to your answer below: does it mean that we can use an image of
> the bones when segmenting our CT images as an additional tissue type?
> Does that have to be part of the dartel procedure or could be perform
> independently?
The new segment was supposed to be able to make a vague stab at handling CT as
well as MR data, but in practice it doesn't do so well. The reason for this
is probably the partial volume effect from having thicker slices. This means
that where the algorithm expects there to be CSF around the brain, instead
there is tissue with an intensity closer to that of bone (ie way higher than
soft tissue or CSF). Because of this, it can't get a good handle on the
intensity distribution of CSF, which causes it to mess up the ventricles.
There is a discretly hidden option to use a non-parametric representation of
the tissue intensity distributions, which I found works slightly better for
CT data (but is still not that great). To enable this, you need to change
the "number of Gaussians" for one or more of the tissue classes to
non-parametric.
In principle, DARTEL does not need to use only GM and WM. You could also use
other tissues (but not all 6, as it implicitly generates a "background"
class).
> BTW, SPM8 has a tiny bug with the edit button in the spm_select procedure.
I'm sure this will get fixed soon - maybe ready for the forthcoming updates.
All the best,
-John
|