[Insight-users] Re: itkHessianRecursiveGaussianImageFilter and Publishing Negative Results.

[Luis Ibanez]

Hi Maisie,

Could you please post a more detailed description
of the behavior that you find to be incorrect in
the Hessian filter ?

We use the Hessian in the application

       InsightApplications/DuctExtractor

that finds tubes in 3D.


You may want to play with this application and attempt
to fine tune the parameters of the tube extraction
algorithm.


About the code below, the Hessian computes second derivatives
along each one of the axis, and smoothing along the remaining
axis. You may want to look at the report by R. Deriche from
INRIA regarding the use of Gaussian Smoothing and computation
of derivatives.


If you still find difficulties using this filter, please post
a short description of your problem, along with the image
that you are trying to segment, to the Insight Journal.

   http://www.insightsoftwareconsortium.org/InsightJournal/


Note that the Insight Journal is an appropriate place for
reporting *NEGATIVE RESULTS*. An activity that is fundamental
in scientific research but that it is ignored by most of
the vanity science Journal where we publish when we need
entries for our CVs. The fact that most journal don't publish
replication of work, and don't publish negative results is
an indication of their decadence and their lack of compliance
with the essential practices of the scientific method.



Your post to the Insight Journal should be done in such a
way that we can replicate what you are doing, (even if it is
the failure to segment a structure). Once we can replicate your
environment, it should be much easier for us to identify the
problem that you are facing and to propose a solution.

The infrastructure of the Insight Journal will allow you to
load large 3D images, source code, etc.

Please let us know if you need assistance for posting your
report to the Insight Journal.


    Best Regards,


      Luis


-----------------
Maisie Wang wrote:
> Hi Luis,
> 
> I have tried to use the itkHessianRecursiveGaussianImageFilter to 
> perform a vesselness measure for MR angiography images. It seems like 
> there could be a bug in:
> 
> void    
> HessianRecursiveGaussianImageFilter<TInputImage,TOutputImage>::GenerateData(void) 
> 
> 
> within the while loop. It seems like the smoothing filter is not 
> performed correctly. Could you explain what the logic is in that bit of 
> code?  I have added the code below.
> 
> 
> while( i < NumberOfSmoothingFilters )
> {
>     while( j < ImageDimension )
>     {
>         if( j != dima && j != dimb )
>         {
>             m_SmoothingFilters[ i ]->SetDirection( j );
>             j++;
>             break;
>         }
>         j++;
>     }
>     i++;
> }
> 
> 
> 
> 
> 
> Thanks,
> Maisie Wang
> 
> 
>