[Insight-users] FW: Problem with Demons Variant Deformable Registration

[Tom Vercauteren]

Hi Sajendra,

I might have talked to fast (again). It seems that the maximum update
step length used in the basic symmetric forces demons implementation
is 1 pixel (and not 0.5 pixels as proposed by Thirion).

Hope this helps,
Tom

On Fri, Apr 11, 2008 at 10:07 AM, Tom Vercauteren
<tom.vercauteren at m4x.org> wrote:
>  Hi Sajendra,
>
>  I thought I fixed the windows issue you mentioned some time ago. Can
>  you try with the current version (revision 6) on the insight journal
>  if it works on your computer? Unfortunately, I don't have access to a
>  windows machine right now.
>
>  The update rule should indeed be '-a1'. I wrote that from memory and
>  got confused...
>
>  What kind of pixel type do you use in your code? There might be some
>  issues with integer types. Did you try using floats or doubles?
>
>  Tell me if you still get large differences using floating point
>  pixels, and the last version of both my code and ITK.
>
>  Tom
>
>  On Thu, Apr 10, 2008 at 10:48 PM, Nithiananthan, Sajendra
>
> <Sajendra.Nithiananthan at rmp.uhn.on.ca> wrote:
>
>
> > Hi Tom,
>  >
>  >  When I noticed the differences I wasn't using your supplied command line
>  >  program but was creating an instance of your implementation of
>  >  itk::FastSymmetricForcesDemonsRegistrationFilter in my own application (I am
>  >  reading data in a different format). I change an include and a typedef to
>  >  compare different implementations, using their default parameters and the
>  >  same sigma for smoothing.
>  >  In ESMDemonsRegistrationFunction the default MaximumUpdateStepLength is 0.5
>  >  and default gradient is Symmetric, which is as you recommend. I don't
>  >  understand your suggestion for the update-rule parameter, "-a2". Looking at
>  >  the command line app, DemonsRegistration.cxx, you need -a1 to create
>  >  itk::FastSymmetricForcesDemonsRegistrationFilter, -a2 would create
>  >  itk::DiffeomorphicDemonsRegistrationFilter. Wouldn't the former act more
>  >  similar to the basic symmetric forces demons implementation? Please correct
>  >  me if I am missing something.
>  >
>  >  As for issues, I did have some difficulties getting the project compiled on
>  >  Windows, but eventually I managed. I think most of that was around the use
>  >  of Getopt in the command line program. I can't recall if there were any
>  >  issues with the registration code itself, if you have tried on Windows you
>  >  would know as well as I.
>  >
>  >  Thank you again,
>  >  Sajendra
>  >
>  >
>  >
>  >  > -----Original Message-----
>  >  > From: tom.vercauteren at gmail.com [mailto:tom.vercauteren at gmail.com]On
>  >  > Subject: Re: FW: Problem with Demons Variant Deformable Registration
>  >  >
>  >  >
>  >  > Hi Sajendra,
>  >  >
>  >  > I am glad to hear that you were able to use my code succesfully.
>  >  >
>  >  > As Torsten pointed out, it should indeed be possible to "get the best
>  >  > of both worlds". In the code I posted on the Insight Journal, It
>  >  > should be possible to use a fast symmetric force that behaves very
>  >  > similarly to the basic symmetric forces demons implementation but has
>  >  > a reduced computation time.
>  >  >
>  >  > The difference you experience might be related to different default
>  >  > parameters. More specifically, from the command line program I posted
>  >  > you should at least put the following options '-l0.5 -a2 -t0' if you
>  >  > want to match the basic symmetric forces demons implementation. Of
>  >  > course you also need to use the same sigmas for smoothing.
>  >  >
>  >  > Another possible issue might be related to the version of ITK that you
>  >  > use. I filed a few bug reports anf forgot when they were fixed:
>  >  > http://www.itk.org/Bug/view.php?id=3985
>  >  > http://www.itk.org/Bug/view.php?id=3958
>  >  > http://www.itk.org/Bug/view.php?id=3973
>  >  > http://www.itk.org/Bug/view.php?id=3972
>  >  >
>  >  > Let me know if you find other issues.
>  >  >
>  >  > Best regards,
>  >  > Tom
>  >  >
>  >
>  >
>  >
>  > This e-mail may contain confidential and/or privileged information for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies. Opinions, conclusions or other information contained in this e-mail may not be that of the organization.
>  >
>  >
>