[Insight-users] How to use itk::SegmentationLevelSetImageFilter

[Joshua Cates]

Hello Yanjun,

1. The input image can be of any scalar data type, it will be converted to
floating point for calculations and output.  The initial front is given by
SetIsoSurfaceValue(f).  For simplicity, the algorithm shifts the input
image values so that the given isosurface f is zero, which is the reason
you see it adding only m_ValueZero to the ActiveLayer (propagating front).


2. The value range for the two thresholds is the min and max of the 
original input image.  This really are just thresholds taken on the input 
image.



Documentation can be found at:

http://www.itk.org/Insight/Doxygen/html/classitk_1_1SegmentationLevelSetImageFilter.html
 
and

http://www.itk.org/Insight/Doxygen/html/classitk_1_1SparseFieldLevelSetImageFilter.html

Also see the Application examples under:

Insight/Applications/LevelSetSegmentation


Josh.


______________________________
 Josh Cates			
 School of Computer Science	
 University of Utah
 Email: cates@sci.utah.edu
 Phone: (801) 587-7697
 URL:   www.cs.utk.edu/~cates


On Fri, 6 Dec 2002, Xu, Yanjun wrote:

> Hi,
>  
> I want to use SegmentationLevelSetImageFilter to do some brain MRI
> segmentation, but didn't make clear how to use it. I have the following
> questions about it.
>  
> 1.	What kind of seed image does it need? Does it accept one label
> (class) or multiple label image? In
> SparseFieldLevelSetImageFilter<TInputImage,
> OutputImage>::ConstructActiveLayer(), it seems that only pixels with
> m_ValueZero  are put into ActiveLayer. Is this ActiveLayer considered
> propagation front? Does it mean that seed label must be zero? Can this seed
> label be point, contour, or area?
>  
> 2.	Threshold SegmentationLevelSetImageFilter , a subclass of
> SegmentationLevelSetImageFilter,  need upperThreshold and lowerThreshold.
> What is the value range for these two thresholds? Are they the thresholds
> for original image or some derived feature image( Curvature, Gradient)? 
>  
> 3.	Does ITK have any filter supporting Multilabel  Fast Marching?
>  
> Thanks a lot.
>  
> Sincerely,
> Yanjun
>  
>