[Insight-users] Narrowband in Reinitiallizelevelset filter

[Sayan Pathak]

Hi Suja,

Not having the images in front of me, I am second guessing your problem based
on our experience. As you may be aware of, constraining front evolution is a
very important issue because of the leakage especially where edge contrast is
poor. While constraining the NarrowBandWidth is one way of controlling the
levelset evolution, researchers have also found shape models to be of great
value. There are several flavors of shape models in the literature. There is
some code in ITK that might be worth taking a look. 

We have found a particular implementation useful in segmenting the corpus
callosum in brain MR images. The code has several components. Please follow
the examples and testing code to navigate the algorithm implementation.

Insight/Examples/Segmentation/GeodesicActiveContourShapePriorLevelSetImageFil
ter.cxx
Insight/Testing/Code/Algorithms/itkGeodesicActiveContourShapePriorLevelSetIma
geFilterTest_2.cxx

Hope this helps, Thanks,
Sayan



-----Original Message-----
From: S, Suja (Research) [mailto:Suja.S at geind.ge.com]
Sent: Wednesday, June 30, 2004 10:32 PM
To: Insight-users at public.kitware.com
Subject: [Insight-users] Narrowband in Reinitiallizelevelset filter


Hi,

I will very much appreciate if you could answer the following question.

 I am using Geodesic Active contour to correct the manual segmented boundary
of hippocampus in MR images.
But it it enlarges the structure too much by taking even the CSF pixels. So I
tried reducing the narrowbanding to 2,
which still took in a lot of wrong pixels. When I checked the code I saw
stopping value as OutputNarrowBandWidth/2 + 2.
What does this mean? Also what are InputNarrowBand and OutputNarrowBand?

If I want the levelset to search only 2 pixels on both sides of input image,
what should I do?

Thanks,
Suja.
_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users