[Insight-users] Level set segmentation parameters to anisotropic images

Wed May 14 09:41:43 EDT 2008

Hi Loic,

You may find useful to read the ITK Software Guide

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

in particular, you should read section 9.3 "Level Set Segmentation"
in pdf-page 563.

As you will see in Figure 9.13 in page 563,
the Level Set is defined as:

       a) f(x,y,z) > 0    is  inside
       a) f(x,y,z) = 0    is  the shape (zero set)
       a) f(x,y,z) < 0    is  outside

   Regards,

      Luis

---------------------
BOURGEOIS Loic wrote:
> Hi,
> 
> Thank you for your very quick reply.
> 
> But I think I put my initial level set in a wrong form because with the same parameters and the initial level set provided by the Fast-Marching filter, the Geodesic filter works (the contour has correctly evolved).
> 
> So I have a simple question : what is exactly the form of the initial level set? (Is it a function f(x,y,z)<0 if (x,y,z) corresponding voxel is inside the object, else f(x,y,z)>0 and f(x,y,z)=0 on the contour? like a signed distance map for example)
> 
> Thanks in advance,
> 
> Loic
> 
> -----Message d'origine-----
> De : Luis Ibanez [mailto:luis.ibanez at kitware.com]
> Envoyé : mardi 13 mai 2008 18:42
> À : BOURGEOIS Loic
> Cc : Insight Users
> Objet : Re: Level set segmentation parameters to anisotropic images
> 
> 
> 
> 
> Hi Loic,
> 
> It will be very hard to track the problem if you just look at
> the full pipeline.
> 
> The best way is to verify every one of the intermediate stages
> of the process.
> 
> A level set segmentation involves a large number of parameters
> and it is hard to figure out which one of them is not set up
> correctly. In general several of them will need to be adjusted.
> 
> 
> Could you please start by connecting an Observer to the filter
> and printing out the iteration events ?
> 
> Also, at the end of execution of the filter you could print
> out the number of iterations that it took to run.
> 
> This will help to determine how much work the filter is
> actually doing.
> 
> E.g. it may be that you are not running enough iterations,
> or it may be that your RMS error tolerance is set too high
> and the filter terminates prematurely...
> 
> 
> 
> You may want to consider playing with SNAP before you continue
> the path of writing your own code. In this way you will learn
> faster about how to fine tune the segmentation parameters.
> 
> 
> 
>      Regards,
> 
> 
> 
>          Luis
> 
> 
> 
> ---------------------
> BOURGEOIS Loic wrote:
> 
>>Hi Luis,
>>
>>Thanks for your suggestion, I have decided to use "GeoInterp: Contour Interpolation with Geodesic Snakes
>>http://www.insight-journal.org/midas/view_item.php?itemid=512" adapted to my problem. I put the distance map (Danielsson) of my pre-segmented object instead of Fast-Marching as the initial level set but the output level set of the geodesic filter is the same of the input.
>>Am I missing something?
>>
>>Regards,
>>
>>Loic
>>
>>-----Message d'origine-----
>>De : Luis Ibanez [mailto:luis.ibanez at kitware.com]
>>Envoyé : lundi 12 mai 2008 18:43
>>À : BOURGEOIS Loic; Insight Users
>>Objet : Re: Level set segmentation parameters to anisotropic images
>>
>>
>>
>>Hi Loic,
>>
>>                    Welcome to ITK !
>>
>>It is hard to tell if the anisotropy of your MET images
>>would prevent LevelSet methods from running well.
>>
>>
>>I would expect that an anisotropy larger than 1:5
>>will make hard for level sets to behave properly.
>>However... this also depends on the actual content
>>of your image...
>>
>>
>>---
>>
>>What I would suggest is for you to try the SNAP application.
>>
>>   You can download binaries from http://www.itksnap.org/
>>
>>or build it from the version in InsightApplications.
>>
>>SNAP is build on top of ITK and provides a nice interface
>>for using LevelSets.
>>
>>Another application that may be useful is Slicer:
>>
>>     http://www.na-mic.org/Slicer/Download/Snapshots/
>>
>>
>>---
>>
>>By testing your images with these applications will get
>>a feeling of whether the anisotropy is too large for the
>>LevelSets to work properly.
>>
>>Please note that ITK LevelSet can be run by turning the
>>consideration of the spacing ON and OFF.
>>
>>You may want to use them with using the spacing ON.
>>
>>Also, a quick experiment that will help you see the
>>influence of anisotropy is to run the GradientMagnitude
>>RecursiveGaussianImagefilter in your image, and then
>>visualizing the image with slice orthogonal to the
>>plan of acquisition.
>>
>>E.g. using a plane XZ, where Z is the direction with
>>the largest spacing. In this way you should see
>>rectangular pixels, and how they may have affected
>>the gradients of the image.
>>
>>
>>Please let us know what you find,
>>
>>
>>
>>     Thanks
>>
>>
>>        Luis
>>
>>
>>----------------------
>>BOURGEOIS Loic wrote:
>>
>>
>>>Hello Mr. Ibanez,
>>>
>>>I am new in this mailing list. I have to segment MET 3D (Microscopy 
>>>Electron Tomography) images which are very anisotropic : contrast is 
>>>quite good along x and y axis but is very low along z axis. So the issue 
>>>is to tune different parameters to the level set (geodesic) between x 
>>>and y axis, and z axis. Is there possible?
>>>
>>>Regards,
>>>
>>>Loic Bourgeois
>>>
>>>__________________________
>>>
>>>Ce message (et toutes ses pièces jointes éventuelles) est confidentiel et établi à l'intention exclusive de ses destinataires. Toute utilisation de ce message non conforme à sa destination, toute diffusion ou toute publication, totale ou partielle, est interdite, sauf autorisation expresse. L'IFP décline toute responsabilité au titre de ce message.
>>>
>>>This message and any attachments (the message) are confidential and intended solely for the addressees. Any unauthorised use or dissemination is prohibited. IFP should not be liable for this message.
>>>
>>>Visitez notre site Web / Visit our web site : http://www.ifp.fr
>>>__________________________
>>
>>__________________________
>>
>>Ce message (et toutes ses pièces jointes éventuelles) est confidentiel et établi à l'intention exclusive de ses destinataires. Toute utilisation de ce message non conforme à sa destination, toute diffusion ou toute publication, totale ou partielle, est interdite, sauf autorisation expresse. L'IFP décline toute responsabilité au titre de ce message.
>>
>>This message and any attachments (the message) are confidential and intended solely for the addressees. Any unauthorised use or dissemination is prohibited. IFP should not be liable for this message.
>>
>>Visitez notre site Web / Visit our web site : http://www.ifp.fr
>>__________________________
>>
> 
> __________________________
> 
> Ce message (et toutes ses pièces jointes éventuelles) est confidentiel et établi à l'intention exclusive de ses destinataires. Toute utilisation de ce message non conforme à sa destination, toute diffusion ou toute publication, totale ou partielle, est interdite, sauf autorisation expresse. L'IFP décline toute responsabilité au titre de ce message.
> 
> This message and any attachments (the message) are confidential and intended solely for the addressees. Any unauthorised use or dissemination is prohibited. IFP should not be liable for this message.
> 
> Visitez notre site Web / Visit our web site : http://www.ifp.fr
> __________________________
> 