[Insight-users] Segmentation issue ( in level set techniques)

Luis Ibanez luis.ibanez at kitware.com
Sat Mar 19 20:29:36 EST 2005


Hi Sotiris,

You may want to combine two techniques:


1) Feature detection
2) Segmentation


For feature detection you may want to build
a coarse model of the bone. A typical candidate
is the itkTubeSpatialObject.  You can do Model
to Image registration as depicted in the ITK
Software Guide


     http://www.itk.org/ItkSoftwareGuide.pdf


Once you get a coarse registration, then you
can use the points in the medial section of the
itkTubeSpatialObject in order to initialize a
LevelSet or a RegionGrowing segmentation.


If you anticipate to receive many images of the
same type, you could also do a "semi-automatic"
segmentation on one of the images, and then use
that result as a model to be automatically fitted
to the new incoming images.



   Regards,


       Luis




----------------------------
Sotiris Dimopoulos wrote:

> Hello to everybody!
> 
> I have a pack of slices of the humerus bone (in dicom
> format) and I want to segment, in each slice, the
> surface inside the bone. The area inside the bone and
> outside the bone has no intensity difference (it is
> black everywhere apart from the pixels that represent
> the bone, which are white). So, imagine a black
> picture that has something that looks like a white
> circle somewhere in the middle.
> 
> I want the procedure to be automatic in each slice (so
> I can’t use stuff like position-seeds) and as a result
> the region growing techniques don’t fit. Moreover, I
> have no intensity difference inside and outside the
> bone, so methods like watersheds (and generally
> intensity-based methods) do not seem appropriate. 
> 
> I try to use "level set segmentation" techniques but I
> can’t find an automatic way of computing the zero
> level set function. Of course, I want to take
> advantage of the edges of the bone and I wish these
> white contours that I have to represent the zero level
> set function. 
> 
> 1) How can I "automatically" extract the initial model
> 
> Ψ(Χ,t=0) that these filters need as an input from each
> dicom slice (feature images)?
>    
> 2) Which filters do you believe will help me solve my
> problem?
> 
> 
>  Many thanks in advance
>           Sotiris
> 
> 
> ____________________________________________________________
> Do You Yahoo!?
> Αποκτήστε τη δωρεάν @yahoo.gr διεύθυνση σας στο http://www.otenet.gr
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
> 





More information about the Insight-users mailing list