[Insight-users] Segmentation data

[Mauro Maiorca]

Hi Christian,

> what I even didn't understand is how to save for example a transformation or
> windowing as a file which describes even the process to get the resulting
> image. Do you unterstand what I would like to get?
>
> Example: I am a physician and I would like to mark a tumor within image data
> of a brain.  I don't want to change the original image data but I would like
> to save a description of how to segment the image for displaying the tumor.

In the first paragraph you are talking about "transformation", broadly
speaking a transformation occurs when you "transform" the original
image into another image. For example, when you rotate an image you
"transform" the original image into a rotated one. You can overwrite
the transformed image to the original one, or you can save the
transformation in another file, depending on what you are trying to
achieve.

Segmentation is a totally different story. Segmentation do not
transform an image into another one, but it is a process of mapping
semantically the "objects" in the image. The output of a segmentation
algorithm might be stored in another file, which -for convenience-
might be an image with the same format/dimension/spacing/etc of the
input image. When you segment an image you end up with the original
image AND another image mapping the original one.

In your example, the segmented image might be an image with the same
dimension/spacing/etc of the original image, but the values of the
segmented image would be all ones in correspondence of the tumour, and
zeros otherwise. Does it make sense? The physician might use a
combination of those two images to have an idea of what goes on, often
a rendering of the segmentation image might help the physician. You
might want to have a look at VTK for the visualization bit.

HTH,
Mauro