Hi Jonathan,
> What is it exactly you are trying to do? It's a bit hard to ever set
> a firm threshold for a tissue class because they are just
> probabilities. If you use a threshold of 0.5, though, you shouldn't
> get voxels that "count" for more than one tissue class. But, it's
> tricky: you could have a voxel that has a .4 probability of being gray
> matter, .4 probability of being white matter, and .2 probability of
> being CSF. Using the 0.5 threshold, this isn't any tissue type, but
> in reality, clearly it is something.
just to chime in, I agree about a (high, such as 0.5) fixed threshold
being hard to defend (although it should be said that this is less of an
issue with the newer segmentation approaches as these tend to produce
values either close to 0 or 1). Excluding low intensity voxels (say,
0.1) in statistics later-on is recommended, but if you want to binarize
segmented images, the approach I still find most convincing is to label
a voxel as belonging to the class for which it has the highest
probability. This avoids setting a fixed threshold in the first place. I
think the VBM toolbox has such a labeling approach integrated, otherwise
it could be scripted.
Cheers,
Marko
--
____________________________________________________
PD Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Universitäts-Kinderklinik
Abt. III (Neuropädiatrie)
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
University Children's Hospital
Dept. III (Pediatric Neurology)
Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
[log in to unmask]
http://www.medizin.uni-tuebingen.de/kinder/epn
____________________________________________________
|