[Insight-users] Special questions regarding GeodesicActiveContourImageFilter

[Oliver Gloger]

Hello ,

I want to segment a MRI –Image and I am trying to use variant Level Set Filters from the ITK. My intention is a refinement of my initial contours. With the LaplacianSegmentationLevelSetImageFilter I got nice results, but some of my segments should grow larger to achieve a congruence with larger gradientMagnitudes in the original image and some others should shrink and split. So I am using the GeodesicActiveContourImageFilter but got segments some have grown very large in comparison with my initial contours. It seems that they do’nt “notice” the important gradientMagnitudes. 
My questions are:

1.)By using the GeodesicActiveContourImageFilter is it necessary to use the sigmoidFilter? I didn’t understand his sense in that context (I know that there is a special description for it). I thought, the GradientMagnitude is my important Input for the GeodesicActiveContourImageFilter. The OutputImage after using the Sigmoidfilter is a white image!!! Could that be my problem? 

2.)	
I have some problems in understanding the initialDistance(==-seedValue) in that context! Should this be the distance from the center point inside my initial contour to the “expected” segment-contour after the segmentation process? I want to get this segment-contour from the Level Set Segmentation, so I can’t know this distance before!!!The SoftwareGuide says: The rule of thumb for the user is to select this value as the distance from the seed points at which the initial contour should be. I have many seed points for one initial segment but don’t really know which distance you mean !!
       
3.)  How can I tell the Level Sets to shrink and not only to expand my initial contours? 
	
Best regards,
Oliver 

-- 
"Feel free" - 5 GB Mailbox, 50 FreeSMS/Monat ...
Jetzt GMX ProMail testen: www.gmx.net/de/go/mailfooter/promail-out