[Insight-users] Re: Question for CurvelevelsetImageFilter ?

[Raul San Jose Estepar]

Hi Zhao,

Could you be more specific about the strange result you are getting?.

By creating isosurfaces corresponding a negative values, you are 
overestimating the object you want to extract. By doing the same with 
positive values you are underestimating the contour  you want to create. 
If the level set is converging properly to the tubular structure you 
want to obtain, the zero level set should be the answer. By generating 
isosurface for -0.05 and 0.05 you are introducing a small offset. In 
case of small vessels you might have a better impression of the 
structure by taking the -0.05 levelset, however this is not the right 
boundary you may want.

You also mentioned that when you create the -0.05 and 0.05 isosurface 
the result seems meaningful. What it is clear is that the isosurface 
corresponding to -0.05 level occludes the other, so you don't need to 
generate the 0.05 one.

/Raul

>Message: 5
>Date: Sat, 31 Jul 2004 16:58:49 +0800
>From: "zhao yong qiang" <zhao-yq at cs.sjtu.edu.cn>
>Subject: [Insight-users] Question for CurvelevelsetImageFilter ?
>To: <insight-users at itk.org>
>Message-ID: <000201c476dc$9b92d330$d500a8c0 at sjtuff2wn2eggg>
>Content-Type: text/plain; charset="us-ascii"
>
>Hi all,
>
>I use the CurveLevelSetImageFilter for MRA segmentation, then I convert
>itkImage(intensity (-4.0---4.0)) to vtkImage. Finally, I use marching
>cube algorithm to display the zero surface, but the result is very
>strange, then I add other two isosurfaces(-0.05 and 0.05), the result
>seemed right.
>
>Can someone explain this?
>
>Thanks in advance!
>
> 
>
>Regards, 
>                          Zhao Yongqiang
>                          Dept. of Computer Science & Engineering
>                          Shanghai JiaoTong University
>                          Shanghai, China
>                          zhao-yq at cs.sjtu.edu.cn
>