[Insight-users] outer surface extraction

[Luis Ibanez]

Hi Ruby,

1) ITK doesn't have many filters for generating triangulations.
    You will find easier to perform the segmentation with ITK and
    then convert the resulting binary mask to VTK in order to use
    VTK contour filters for surface extraction.

2) You will find the code examples used for segmenting the
    Visible Woman dataset under


     InsightDocuments/SoftwareGuide/Cover/Source


3) You can also use the ITK plugins in VolView. In particular
    ThresholdLevelSet, and confidence connnected run for
    growing a region on the background of the image instead of
    the tissue.

    You can download the free version of VolView from
    http://www.kitware.com/products/volview.html

    It already has the binary version of the plugins.
    You can download the source code of the plugins from

      InsightApplications/VolviewPlugins


4) You can also try the following segmentation applications
    in InsightApplications:

     - FastMarchingLevelSet
     - ShapeDetectionLevelSet
     - SimpleLevelSetsExample
     - ThresholdSegmentationLevelSet
     - ThresholdSegmentationLevelSetFltkGui

    Remember that for segmenting the skin, the easy way to go
    is to segment the 'complement' of the head. That is, put
    seed points outside the head and grow a region representing
    the background of the head. Once you get this mask, you can
    simply negate it and you will have a segmentation of the
    head/skin.



5) If you data is too noisy you should first experiment with
    denoising filters before applying registration filters.
    Here are some filters that you want to try in your data:

     - MedianImageFilter
     - CurvatureAnisotropicDiffusionImageFilter
     - GradientAnisotropicDiffusionImageFilter
     - CurvatureFlowImageFilter
     - BilateralImageFilter



Regards,


    Luis



-------------------
Ruby Shamir wrote:

> Hi All,
> 
> I am trying to get the 3D outer surface of the head from MRI scans.
> 
> I have a 3D (very noisy) image of the head.
> 
> To do so we tried 3 approaches:
> 
> 1. Using VTK we cleaned the data and "shot" rays during bounding box and
> saved the closest intersection points.
>    The cloud of points look pretty good. But we need a "good"
> triangulation for curvature calculation. (Delauny is not good for that
> for example).
> 
> 2. Using MATLAB we segmented the outer surface slice by slice
> (automatic). However it is important to do this on 3D data since the
> result is noisy in the "z" direction (we smoothed it and it is till not
> good enough).
> 
> 3. Therefore we are seeking for a 3D segmentation approach; threshold
> level set as Luis suggested us seems good. 
> 
> My questions are:
> 
> 1. Does someone know a consistent triangulation method/implementation?
> 
> 2. Does someone have a 3D segmentation of the outer surface examples
> (with complex objects and noisy data)?
> 
> Thank you,
> 
> (and sory for the long letter)
> 
>       Ruby
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>