[Insight-users] Segmentation

[Luis Ibanez]

Hi Suresh,

Currently there is not an algorithm in
ITK that will perform exactly this.

How much anatomical detail are you considering
for brain tissues ?

For example: will { White matter, Gray matter, CSF }
be enough ?

or are you looking to get specific brain structures ?

If the three large classes are enough, it will be
easy to put together a filter based on the Flood
Fill iterator and a connection criteria.

If you want to get detailed anatomical structures,
a good method will be to do Segmentation by Registering
the patient's brain with a presegmented brain atlas.
This can be done with the Demon's algorithm implemented
in ITK. An example of this approach can be found in:

      Insight/Examples/ISBRValidation

This is a full validation study developed by Lydia Ng
at Insighful.


Please let us know what approach you find more reasonable
for your application.


    Luis

==============================================

suresh wrote:
> Hi Luis,
> 
> Thanks for you help on Connectedness.
> What i'm trying to do is findout all 26-connected components in a 
> volume, and keep largest of them.This way i want segment the brain 
> tissues from the non-brain.
> 
> Is there any algorithm implemented ITK for this
> Thank you.
> 
> suresh
> _______________________________________________
> Insight-users mailing list
> Insight-users@public.kitware.com
> http://public.kitware.com/mailman/listinfo/insight-users
>