[Insight-users] Results of Demons Deformable Registration

[Lydia Ng]

Hi Hari,

We had some issues with the demons not behaving well in some cases when =
the
spacing is anisotropic. This is due to the differences in units in =
Thirion's
original equation which adds a gradient^2 to an intensity^2 term. =
Recently,
Luis Ibanez came up with a better way of normalizing than in the =
previous
implementation. The changes were made after the ITK 1.6 (we'll tried to =
patch
the 1.6 release with these changes).

- Lydia

> -----Original Message-----
> From: Hari sundar [mailto:hsundar at seas.upenn.edu]
> Sent: Thursday, February 19, 2004 4:07 PM
> To: Lydia Ng
> Subject: RE: [Insight-users] Results of Demons Deformable Registration
>=20
> Hi Lydia,
>=20
> I ran the registration again using the class that you provided, and =
now
> it works !!! ...
>=20
> What were the changes ?
>=20
> thanks for the help.
> -Hari
>=20
> On Thu, 2004-02-19 at 17:58, Lydia Ng wrote:
> > Hi Hari,
> >
> > What are the spacing values of your images?
> > Could you do an experiment for me? Can you repeat your registration
> using
> > either the cvs ITK or the updated DemonsRegistrationFunction class
> attached?
> >
> > Thanks,
> > Lydia
> >
> > > -----Original Message-----
> > > From: Hari sundar [mailto:hsundar at seas.upenn.edu]
> > > Sent: Thursday, February 19, 2004 2:36 PM
> > > To: Lydia Ng
> > > Subject: RE: [Insight-users] Results of Demons Deformable =
Registration
> > >
> > > Hi Lydia,
> > >
> > > I am using ITK 1.6. I am using it straight, this is not in a =
multi-
> scale
> > > framework. Will that make a difference ?
> > >
> > > The other image looks almost the same and has the same =
size/intensity
> > > ranges (same subject, same machine, same modality). Therefore, I =
did
> not
> > > use the histogram matching presprocessing.
> > >
> > > It's the simplest possible pipeline. I don't know why the images =
look
> > > corrupted.
> > >
> > > thanks,
> > > -Hari
> > >
> > > ------------------------------------------------------
> > > The Pipeline
> > > ------------------------------------------------------
> > > fixedImageReader->SetFileName( fixed );
> > > movingImageReader->SetFileName( moving );
> > >
> > > fixedImageCaster->SetInput( fixedImageReader->GetOutput() );
> > > movingImageCaster->SetInput( movingImageReader->GetOutput() );
> > >
> > > filter->SetFixedImage( fixedImageCaster->GetOutput() );
> > > filter->SetMovingImage( movingImageCaster->GetOutput() );
> > >
> > > filter->SetNumberOfIterations( 150 );
> > > filter->SetStandardDeviations( 5.0 );
> > >
> > > filter->Update();
> > >
> > >
> > > fieldWriter->SetInput( filter->GetOutput() );
> > > fieldWriter->SetFileName(defs);
> > >
> > > fieldWriter->Update();
> > >
> > > FixedImageType::Pointer fixedImage =3D =
fixedImageReader->GetOutput();
> > >
> > > warper->SetInput( movingImageReader->GetOutput() );
> > > warper->SetInterpolator( interpolator );
> > > warper->SetOutputSpacing( fixedImage->GetSpacing() );
> > > warper->SetOutputOrigin( fixedImage->GetOrigin() );
> > > warper->SetDeformationField( filter->GetOutput() );
> > >
> > > writer->SetFileName( warp );
> > >
> > > caster->SetInput( warper->GetOutput() );
> > > writer->SetInput( caster->GetOutput()   );
> > > writer->Update();
> > >
> > > ---------------------------------------------------------------
> > > On Thu, 2004-02-19 at 16:40, Lydia Ng wrote:
> > > > Hari,
> > > >
> > > > Which version of ITK are you using?
> > > > Are you using demons straight or in a multi-scale framework?
> > > > You only posted one image (orig.jpg) what does the other image =
look
> > > like?
> > > > Did you do any histogram matching preprocessing before hand? If =
so,
> have
> > > you
> > > > check if the output looks ok?
> > > >
> > > > - Lydia
> > > >
> > > > > -----Original Message-----
> > > > > From: Hari sundar [mailto:hsundar at seas.upenn.edu]
> > > > > Sent: Thursday, February 19, 2004 1:08 PM
> > > > > To: Insight-users at itk.org
> > > > > Subject: [Insight-users] Results of Demons Deformable =
Registration
> > > > >
> > > > > Hi,
> > > > >
> > > > > I tried to use the demons deformable registration example
> (modified)
> > > to
> > > > > register two 3D MR datasets. The registration results seem to =
be
> > > culled
> > > > > at some intensity level. Please see the attched images.
> > > > >
> > > > > Why is this happening ?
> > > > >
> > > > > thanks,
> > > > > -Hari
> > > > >
> > > >
> >