[Insight-users] How to use GeodesicActiveContour with 3D images

[Luis Ibanez]

Hi Anders,

This application is generic in nature. It is not
intended to be used with a particular type of images.

You could use for example the brain images from the
brainweb project available at the "Data" link of
http://www . itk . org

Applications that read images using the ImageFileReader
do not need to set up spacing, nor size explicitly in
the code.  This information is obtained directly from
the reader. Otherwise you will end up in the situation
where your application will only be able to read images
of a specific fixed size.

The difference between the 2D and 3D versions is mostly
related to the user interaction capabilities and the
image display functionalities.  The internal ITK code
is exactly the same in both 2D and 3D.

We have used the Geodesic active contours method for
segmenting abdominal aortic aneurisms. You may want
to try the ITK plugins for Volview. You can download
the free version of VolView from:

   http://www . kitware . com/products/volview . html


The segmentation can be done directly in 3D. There is no
need for doing slice by slice processing. Slice base
processing is in general a hack of the past. A remanent
of the time where fully N-dimensional methods were not
available.

About the seed points in this application, they are
denoting pixel-indices instead of physical coordinates
in millimeters. Therefore Z is a slice number.

You don't need to worry about this detail though.
Simply select your seed points by clicking on the
displayed image.



Regards,


   Luis


---------------------
Anders Almås wrote:
> Hi!
> 
> I am looking at the GeodesicActiveContourImageFilter in the
> demo-application. I have seen that it should be able to deal with 3D
> images. And my first question is:
> Is there any particular 3D images that are intended to use? because I
> don't see where they set some parameters about the size and spacing of the
> image(for example for *.raw images)
> And is there much to change in the 2D version of
> GeodesicActiveContourImageFilter? I use that one to a series of 2D images
> in png format. My goal is to use those images segmented in 2d as feature
> images for 3d segmentation. My segmentation problem is about segmenting
> aortic aneurysm..
> 
> The final question is about the seed points in the demo application:
> The z coordinate, is that just denoting the image slice? or is it a
> coordinate as x and y?
> 
> -------------------------------------------------------------------------------
> Navn:   Anders Almås
> E-post: andersal at stud . ntnu . no
> Tlf:    90550374
> -------------------------------------------------------------------------------
> 
> 
>