[Insight-users] About Level sets

[Luis Ibanez]

Hi Rashed,

The speed image is a grayscale image, typically with pixel
values in the range from 0.0 to 1.0.  The image must have
values 0.0 in the places where you want the LevelSet front
to stop, and should have values close to 1.0 in the places
where you want to levelset front to propagate.

You will find multiple examples of how a good speed image
should look like in the ITK Software Guide:


       http://www.itk.org/ItkSoftwareGuide.pdf


See for example, section 9.3.1 "Fast Marching Segmentation",
in pdf-pages 565-573. You will see examples of speed images
in figures 9.16, 9.18 and 9.22. All of these illustrate
speed images based on the gradient magnitude.



     Regards,


        Luis



-------------------
rashedk wrote:
> Thanks for your reply Luis, and I am glad to know that this is possible using
> ITK. 
> 
> But I dont understand what I need to put into this *speed image*. Is this a
> binary image or grey-scale? Can you give me an example of what a speed image
> would contain if the propagating front speed was governed by the image's
> intensity gradient magnitude?  
> 
> Thanks for your reply, I really appreciate. 
> 
> Regards,
> Rashed karim 
> 
> 
> 
> Luis Ibanez wrote:
> 
>>
>>Hi Rashed,
>>
>>                     Yes, it is possible.
>>
>>
>>Note that, for example, in FastMarching, *you* build the
>>speed image. Therefore, you can simply precompute the images:
>>
>>    a) Gradient magnitude
>>    b) Vesselness
>>
>>and combine them in any way you want to generate the speed image.
>>It seems, for example, that a multiplication of the two images will
>>produce the "AND" effect that you are looking for.
>>
>>
>>Then, finally, pass the computed speed image to the FastMarching
>>filter, as you have already seen it described in the ITK Software
>>Guide.
>>
>>
>>If you want to make a more sophisticated mixture of terms,
>>then you can easily write your own LevelSet filter, by still
>>reusing the ITK framework. In this case, you have to write
>>two classes:  X_LevelSetImageFunction and X_LevelSetImageFilter.
>>In this scenario, you will find useful to look at the code
>>implementing other LevelSet filters.
>>
>>
>>Also, since you are interested in Vessel segmentation, you may
>>find useful the following paper posted in the Insight Journal:
>>
>>
>>"Generalizing vesselness with respect to dimensionality and shape"
>>by Antiga, Luca
>>http://www.insight-journal.org/InsightJournalManager/view_reviews.php?back=search.php%3Ftexte%3Dvessel&journalid=9&pubid=175
>>
>>	
>>"Vessel Enhancing Diffusion Filter"
>>Enquobahrie, Andinet; Ibanez, Luis; Bullitt, Elizabeth; Aylward, Stephen
>>http://www.insight-journal.org/InsightJournalManager/view_reviews.php?back=search.php%3Ftexte%3Dvessel&journalid=8&pubid=163
>>
>>
>>
>>If you find these papers useful, you are strongly encouraged to
>>support OPEN ACCESS by posting reviews to the papers. Please note
>>that the reviews of the Insight Journal are not intended to be the
>>empty reviews of decadent traditional journals where the reviewers
>>simply speculate on what "they think" about the paper. On the contrary
>>the Insight Journal expects comments from fellow developers of image
>>analysis software who *actually* compiled and run the source code that
>>is accompanying the paper, and report on their experience on the use
>>of the code.
>>
>>
>>
>>    Peer-review without reproducibility is mindless babbling.
>>
>>    We already have had enough of that in our field...
>>    It is time to raise from the muddy waters of narcissism
>>    and move our field up to the standards of scientific work.
>>
>>
>>
>>By the same token, If you find a way of integrating the two images
>>as input to your level set filter, you are also *strongly encouraged*
>>to submit your work as a paper to the Insight Journal. Here again,
>>papers to the Insight Journal are not about the childish quest for
>>"originality" claims that disgustingly plague the decadent journals
>>of our field. Instead, we expect your paper to demonstrate that the
>>practical solution that you found to your real image processing
>>problem, actually works. The demonstration is provided not by textual
>>arguments in your paper, but by the source code, parameters, images,
>>and instructions that you provide, describing how to *REPRODUCE* your
>>tests.
>>
>>
>>  Please let us know if you have any further questions,
>>
>>
>>     Thanks
>>
>>
>>        Luis
>>
>>
>>
>>------------------
>>rashed karim wrote:
>>
>>>Hi Everyone,
>>>
>>>Is it possible to use the ITK level set implementations (such as Fast 
>>>marching, Geodesic, Laplacian, etc.) to drive a propogating front whose 
>>>speeds are governed by let's say two factors: 1. Gradient magnitude 2. 
>>>Frangi's vesselness. I want the propagating front to stop speeding at 
>>>points which give high responses to the vesselness filter AND have high 
>>>image gradient magnitudes.
>>>
>>>Does ITK level set implementations allow users to define their own 
>>>criterion which limits the speed of the active contour? I have had a 
>>>good look at the ITK software guide, and it seemed like this isn't 
>>>possible.
>>>
>>>Thanks in advance,
>>>
>>>Regards,
>>>Rashed Karim
>>>
>>>
>>>------------------------------------------------------------------------
>>>
>>>_______________________________________________
>>>Insight-users mailing list
>>>Insight-users at itk.org
>>>http://www.itk.org/mailman/listinfo/insight-users
>>
>>_______________________________________________
>>Insight-users mailing list
>>Insight-users at itk.org
>>http://www.itk.org/mailman/listinfo/insight-users
>>
>>
> 
>