[ITK Community] Smoothness term in ITK

[Luis Ibanez]

Hi Hana,

The image registration framework in ITK does
not provide verification of the solutions per se.

This is something that would have to be done
at the application level.

One could, of course, use ITK components to
implement such verification.  But still, it requires
a separate implementation.

Note that there is no "truth" in the output of a
registration process.

There are only solutions that are:

       "good enough for  the constrains
              of a given application".



   HTH


       Luis




On Sat, Feb 8, 2014 at 10:43 AM, Hana Cherif <hanacherif62 at gmail.com> wrote:

> Hello all ,,
>
> Another question please about the smoothness term (regularity term )
>
> We know all that the image registration's methods implimented in ITK based
> on a variational approache ,, wich serves to minimize an energy includes
> two terms : similarity term wich measure similarities of the two images,
> and a regularity term wich ensure the regularity of that transformation.
> For me this two terms are essentiel, mainly  in the processus of deformable
> registration !! My question is how did the regularity of the solution is
> verified in the model proposed in ITK ,, i mean the BSpline registration
> (DeformableRegistration4)
>
> Thanks in advance!!
>
> Hana
>
>
>
>
> 2014-01-21 6:48 GMT+01:00 Hana Cherif <hanacherif62 at gmail.com>:
>
> Thank you so much Matt,, Luis ,,  David,,  thank you all,,
>>
>>
>> 2014/1/20 Matt McCormick <matt.mccormick at kitware.com>
>>
>>> Hi,
>>>
>>> There is a class implemented that does Gaussian smoothing on a
>>> displacement field transform during a registration iteration update:
>>>
>>>
>>> http://itk.org/Doxygen/html/classitk_1_1GaussianSmoothingOnUpdateDisplacementFieldTransform.html#a486d7a5466744497b3793b8161472fdc
>>>
>>> Hope this helps,
>>> Matt
>>>
>>> On Sun, Jan 19, 2014 at 10:43 PM, David R. Haynor
>>> <haynor at u.washington.edu> wrote:
>>> > hi all,
>>> >
>>> > a smoothing term is easily incorporated into registration, along with
>>> its
>>> > gradients.  i don't understand luis's comment.
>>> >
>>> > -dh
>>> >
>>> >
>>> > On Mon, 20 Jan 2014, Andras Lasso wrote:
>>> >
>>> >>
>>> >> It would be great to have a term in the registration metric to
>>> penalize
>>> >> non-smoothness of the BSpline (displacement of the control points).
>>> >> Currently, the
>>> >> BSpline registration in ITK cannot be used on many radiation therapy
>>> CT
>>> >> scans, because the BSpline control points wander randomly in the
>>> soft-tissue
>>> >> regions
>>> >> that look quite homogeneous in CT. For example, the BSpline
>>> registration
>>> >> implementation in Plastimatch can penalize the control point
>>> displacement
>>> >> and the
>>> >> result is a nice smooth displacement field.
>>> >>
>>> >>
>>> >>
>>> >> Andras
>>> >>
>>> >>
>>> >>
>>> >> From: Community [mailto:community-bounces at itk.org] On Behalf Of Luis
>>> >> Ibanez
>>> >> Sent: January 19, 2014 3:44 PM
>>> >> To: Hana Cherif
>>> >> Cc: community at itk.org
>>> >> Subject: Re: [ITK Community] Smoothness term in ITK
>>> >>
>>> >>
>>> >>
>>> >> Hana,
>>> >>
>>> >>
>>> >>
>>> >> A smoothing term, will probably only make sense for the Deformable
>>> >> transforms:
>>> >>
>>> >>
>>> >>
>>> >> * Kernel Transform
>>> >>
>>> >> * Displacement Field
>>> >>
>>> >> * BSpline Deformations
>>> >>
>>> >>
>>> >>
>>> >> Note that, this would be rather a measure that is computed after the
>>> fact,
>>> >>
>>> >> rather than something that is computed during the registration
>>> process.
>>> >>
>>> >>
>>> >>
>>> >> Although in the case of the Demons registration, there is indeed the
>>> >> equivalent
>>> >>
>>> >> of a smoothness constraint.
>>> >>
>>> >>
>>> >>
>>> >>    Regards,
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>        Luis
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >> On Sun, Jan 19, 2014 at 3:27 PM, Hana Cherif <hanacherif62 at gmail.com>
>>> >> wrote:
>>> >>
>>> >> Hi Mr.  Louis
>>> >>
>>> >> I  refer to transform classes in itk ,, thanks
>>> >>
>>> >>
>>> >>
>>> >> 2014/1/19 Luis Ibanez <luis.ibanez at kitware.com>
>>> >>
>>> >> Hi Hana,
>>> >>
>>> >>
>>> >>
>>> >> Are you referring to the itk::Transform classes ?
>>> >>
>>> >>
>>> >>
>>> >> or to Level Set filters ?
>>> >>
>>> >>
>>> >>
>>> >> Could you please elaborate a bit more ?
>>> >>
>>> >>
>>> >>
>>> >>     Thanks
>>> >>
>>> >>
>>> >>
>>> >>          Luis
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >> On Sun, Jan 19, 2014 at 3:21 PM, Hana Cherif <hanacherif62 at gmail.com>
>>> >> wrote:
>>> >>
>>> >>       Hello,, all!!
>>> >>
>>> >> A question please ::  In addition to the similarity term,,  is there a
>>> >> componnent  in itk used to compute the smothness term of that
>>> transformation
>>> >> ??
>>> >>
>>> >> Thanks for help in advance!!
>>> >>
>>> >>
>>> >>
>>> >> _______________________________________________
>>> >> Community mailing list
>>> >> Community at itk.org
>>> >> http://public.kitware.com/cgi-bin/mailman/listinfo/community
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >
>>> > _______________________________________________
>>> > Community mailing list
>>> > Community at itk.org
>>> > http://public.kitware.com/cgi-bin/mailman/listinfo/community
>>> >
>>> _______________________________________________
>>> Community mailing list
>>> Community at itk.org
>>> http://public.kitware.com/cgi-bin/mailman/listinfo/community
>>>
>>
>>
>
> _______________________________________________
> Community mailing list
> Community at itk.org
> http://public.kitware.com/cgi-bin/mailman/listinfo/community
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/community/attachments/20140210/a33fd934/attachment-0002.html>