[Insight-developers] changes in level set routines since ITK 3.4 breaks itk-snap

[Luca Antiga]

Found the problem. As expected, it was my fault.

itkParallelSparseFieldLevelSetImageFilter.txx line 2526:

   const ValueType inside_value  = (max_layer+1) *  
m_ConstantGradientValue;
   const ValueType outside_value = -(max_layer+1) *  
m_ConstantGradientValue;

should be replaced with

   const ValueType outside_value  = (max_layer+1) *  
m_ConstantGradientValue;
   const ValueType inside_value = -(max_layer+1) *  
m_ConstantGradientValue;

Now, I kick myself in the foot for not testing the parallel filter  
enough.
I looked at the test for the ParallelSparseField filter and it's  
basically not much more than a
coverage test, as there is no comparison with a baseline image.  
Probably this is a thing that
should be fixed as well.

Now, the real problem is that the latest releases don't work with  
Snap, and that parallel
sparse field filter is basically useless the way it is in the last  
two releases.

In the meantime, should I check the fix in and close the bug?


Luca



On Aug 6, 2008, at 1:43 AM, Casey Goodlett wrote:

> Hi Luis,
>
> I submitted a bug report in mantis with the requested information  
> here http://www.itk.org/Bug/view.php?id=7454
>
> Thanks
>
> Casey
>
> On Tue, Aug 5, 2008 at 5:21 PM, Luis Ibanez  
> <luis.ibanez at kitware.com> wrote:
>
> Hi Casey,
>
> Could you please give us a hand by providing a description
> of the circumstances in which this error becomes visible ?
>
> E.g. a step-by-step set of instructions will be ideal.
>
> We should then be able to introduce in ITK a new Nightly
> test that exercise this same case, and then we can proceed
> to fix the problem.
>
>
>  Thanks for any advice,
>
>
>      Luis
>
>
> ---------------------
>
> Bill Lorensen wrote:
> Casey,
>
> Snap is an important ITK customer. We'll work with you to figure  
> out the issues.
>
> Bill
>
> On Tue, Aug 5, 2008 at 5:15 PM, Casey Goodlett  
> <gcasey at sci.utah.edu> wrote:
>
> It's a functionality change.
>
> On Tue, Aug 5, 2008 at 3:14 PM, Bill Lorensen  
> <bill.lorensen at gmail.com>
> wrote:
>
> Is this a compilation error or functionality change?
>
> On Tue, Aug 5, 2008 at 5:06 PM, Casey Goodlett <gcasey at sci.utah.edu>
> wrote:
>
> Hi,
>
> I have recently compiled itk-snap 1.6.0.1 against the latest  
> version of
> ITK,
> and the automatic segmentation routines in SNAP no longer behave
> correctly.
> When initializing the segmentation routine with a bubble, the bubble
> does
> not get correctly converted into a label map using newer versions of
> ITK,
> and the level set evolution does not behave correctly.  Going back to
> ITK
> 3.4 fixes the problem.
>
> I've noticed that there were changes made to the level set routines in
> ITK
> with respect to bug 0006156 (
> http://public.kitware.com/Bug/view.php?id=6156
> ).  I've checked out CVS of ITK before and after these changes and can
> confirm that the problem came about in ITK between 2008-03-01 and
> 2008-03-06.
>
> I tried toggling the ITK_DEPRECATED_LEVELSET_INTERPOLATION flag in the
> cmake
> config but there was no effect on the problem.
>
> I'm unsure how to solve the problem so that snap can be used  
> against the
> current version of ITK.  Can someone explain how the behavior of the
> level
> set filters changed?
>
> Thank you
>
> --
> Casey Goodlett
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers
>
>
>
>
>
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers