[Insight-users] level set pipeline

[Luis Ibanez]

Hi Ilker,

Yes,
you could try a Level Set Pipeline like this one.

Note however that when you connect the feature image to the
ThresholdSegmentation Level Set filter what it will do with
it is to compute its laplacian and multiply it by a factor
(that you must provide) "EdgeWeight" and will add it to the
threshold contribution in order to compute a final speed image.

You may want to look at the code in:


       Insight/Code/Algorithms
           itkThresholdSegmentationLevelSetFunction.txx

in lines 31-88 in order to get familiar with the actual operation
of the filter.



Take into account that you will have *a lot* of intermediate
parameters to fine tune on each of the filters involved.

I would strongly suggest to to go step by step, fine tunning
the parameters of every stage of your pipeline before you add
the next stage.


For example

a) Make sure that the output of the Sigmoid looks almost like
    a binary image, before you attempt to connect the output to
    the FastMarching filter.

b) Make sure that you select many seeds close to the medial line
    of the object to be segmented, and you pass them to the Fast
    Marching filter.  Make sure that the output of Fast Marching
    is satisfactory before you attempt to pass it to the Threshold
    segmentation filter.

c) Fine tune first the threshold values of the Threshold segmentation
    image filter, and then adjust the value of the EdgeWeight parameter.
    Make sure that you get a reasonable segmentation before you pass it
    to the Geodesic active contour filter.

d) Fine tune the three weights of the GeodesicActiveContour filter.
    If you did a good job in steps (a-c) you will not need a lot of
    propagation weight, because the level set should already be quite
    close to the borders of your image. So probably you could focus
    on the Advection weight and the Curvature weight.


Note: You *really* need an intermediate visualizatoin, different from
just writing images to files. If you don't use an intermediate
visualization It will take you *a long* time to fine tune all those
twenty or so parameters in your pipeline.


  Regards,


    Luis





--------------------------
ilker hacıhaliloğlu wrote:
> hi all
>  
> i am using the level segmentation methods i am just wondering if i can 
> construct a pipeline like this one ??
>  
>  
> 
> smoothing->SetInput( reader->GetOutput() );
> 
> gradientMagnitude->SetInput( smoothing->GetOutput() );
> 
> sigmoid->SetInput( gradientMagnitude->GetOutput() );
> 
> fastMarching->SetInput( sigmoid->GetOutput() );
> 
> thresholdSegmentation->SetInput( fastMarching->GetOutput() );
> 
> thresholdSegmentation->SetFeatureImage( reader->GetOutput() );
> 
> geodesicActiveContour->SetInput( thresholdSegmentation->GetOutput() );
> 
> geodesicActiveContour->SetFeatureImage( sigmoid->GetOutput() );
> 
> thresholder->SetInput( geodesicActiveContour->GetOutput() );
> 
> writer->SetInput( thresholder->GetOutput() );
> 
> 
> 
> -- 
> ------------------------------------------------------------------
> 
> '.....ich möchte mir eine ader öffnen, die mir die ewige freiheit 
> schaffte....'
> 
> --------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users