[Insight-users] Error on watershed example (from ITK1.8\Examples\WatershedSegmentation1.cxx

[neha k]

Hello,

Can I use watershed example (from ITK1.8\Examples\WatershedSegmentation1.cxx) with Brain MRIs (Raw format)?  When I try to use Watershed example on Brain MRI with tumour, with following parameters range-  

Conductance Term -> 1-4, No. of Iterations - 3 - 10, Threshold (%) - 0.0 - 0.3, Level Of Watershed 0.01 - 0.05,
Principle Componenet Analysys - On. I am not sure if I need to use Principle Componenet Analysys.

I get following error: ------>

itk::ExceptionObject (0104F288)
Location: "Unknown"
File: C:\ITK_1.8\InsightToolkit-1.8.0\Code\Algorithms\itkWatershedSegmentTreeGen
erator.txx
Line: 436
Description: itk::ERROR: itk::watershed::SegmentTreeGenerator::MergeSegments:: A
n unexpected and fatal error has occurred. This is probably the result of overth
resholding of the input image.

Do you have any suggestions?  I have attached the image.  I give .mhd file as an input to that example program.

Thanks.



Luis Ibanez <luis.ibanez at kitware.com> wrote:
Hi Neha,

We don't have examples of these two filters are this point.
This issue has been entered as a feature request in the
bug tracker. Bug # 1309:

http://www.itk.org/Bug/bug.php?op=show&bugid=1309&pos=0

----

Regarding your question about pre-post contrast:

Yes, it is worth to give it a try at subtracting these two
images. You will recognize if registration is needed if
you see a lot of borders in the structures after the
subtraction.

It is quite likely that you will need deformable registration
since the image acquisition process for Breast under MRI does
not restrict patient movement.

The deformable registration methods in ITK have been previously
applied to registration of MRI Breast images. In particular the
combination of BSplineDeformableTransform and MutualInformation.

Look for example at the Tutorials

http://www.itk.org/HTML/Tutorials.htm

In particular the session on Registration by Dr. Lydia Ng.
http://www.itk.org/CourseWare/Training/RegistrationMethodsOverview.pdf

You will find examples on deformable registration of pre/post
constrast MRI Breast images in page 64.



Regards,



Luis



--------------------------
neha k wrote:

> Hello Luis,
> Thanks for your reply. Is there any example on 
> GrayScaleFillHoleImageFiler or GrayScaleGrindPeakImageFilter? What 
> book can I refer to understand these filters?
> 
> Also, can you please let me know what you mean by "fine tune the structuring
> element for matching the relative size of calcifications in image". 
> 
> Since I need pair of breast MRIs (Which unfortunately are not avilable 
> to me) I checked mypacs.net, where I found one set of images GRE-images 
> at following link - 
> http://www.mypacs.net/cgi-bin/repos/mpv3_repo/wrm/repo-view.pl?cx_subject=28078&cx_image_only_mode=off&cx_repo=mpv3_repo&cx_from_folder 
> =
> Fil. 9,10 and 11. Does it make sense to use these 2 images, 
> subtracting them and applying Watershed on the result. If so, how can I 
> decide by checking these images if that needs deformable registration?
> Any input on this is most appreciated. 
> Thanks,
> Neha
> 
> */Luis Ibanez /* wrote:
> 
> 
> HI Neha,
> 
> You can find Breat images at
> 
> www.mypacs.net
> 
> you can create an account for free and
> download real datasets.
> 
> Look for the title "Shared Cases" and
> click in "Breast".
> 
> --
> 
> You may want try the advanced Mathematical Morphology
> methods such as: FillHole and GrindPeak.
> 
> http://www.itk.org/Insight/Doxygen/html/classitk_1_1GrayscaleFillholeImageFilter.html
> http://www.itk.org/Insight/Doxygen/html/classitk_1_1GrayscaleGrindPeakImageFilter.html
> 
> Taking an input imag, applying one of these filter and
> then subtracting from the original may enhance spiculations
> and calcificaitions. You will have to fine tune the structuring
> element for matching the relative size of calcifications in your
> image.
> 
> 
> Regards,
> 
> 
> Luis
> 
> ----------------------------
> neha k wrote:
> 
> > Hello All,
> >
> > I am working on Breast Image Segmention for deleneating lesions,
> ducts
> > and fat tissues. Problem is that I don't have Breast Images with
> > contrast agent used. Hence I can't have pre and post contrast agent
> > to subtract them and locate tumour etc. What will be the best way to
> > segment such breast images (in RAW format). I am currently trying to
> > use Watershed segmentation and not getting any good results with it.
> > It causes lot of oversegmentation. I have not used Level Set Seg.
> > method yet.
> > Any input is appreciated.
> >
> > Thanks,
> > Neha
> >
> >
> 
> 
> 
> ------------------------------------------------------------------------
> Do you Yahoo!?
> Yahoo! Mail – CNET Editors' Choice 2004. Tell them what you think 
> . 
> 






__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20041109/4f2e6f77/attachment.html