[elastix] 3D Bspline registration (Stefan Klein)

Kristen Tgavalekos kristen.tgav at gmail.com
Mon Aug 15 16:46:36 CEST 2011


Hi Stefan,

Thank you very much for your quick response. I set the default pixel value
to 500 and those holes did indeed have a value of 500.

Is there a way to specify that data in certain slices should always remain
in the same plane while data in other slices are allowed to deform in any
direction? MovingImageDerivativeScales seems like it might help if there is
a way to specify that parameter for specific parts of the data.

Alternatively, is there a way to specify fixed image landmark positions AND
moving image landmark positions as inputs for registration? I was looking at
the thin-plate splines transform which appears to only allow fixed landmarks
as input. Ideally, I would want to continue using a BSpline transform but be
able to provide these landmarks in order to help guide the registration.

Thank you,
Kristen

On Thu, Aug 11, 2011 at 6:00 AM, <elastix-request at bigr.nl> wrote:

> Send elastix mailing list submissions to
>        elastix at bigr.nl
>
> To subscribe or unsubscribe via the World Wide Web, visit
>        http://lists.bigr.nl/mailman/listinfo/elastix
> or, via email, send a message with subject or body 'help' to
>        elastix-request at bigr.nl
>
> You can reach the person managing the list at
>        elastix-owner at bigr.nl
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of elastix digest..."
>
>
> Today's Topics:
>
>   1. Re: Elastix Ubuntu New User (Stefan Klein)
>   2. Re: Elastix Ubuntu New User (Robinson, Jamie)
>   3. 3D Bspline registration (Kristen Tgavalekos)
>   4. Re: 3D Bspline registration (Stefan Klein)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Wed, 10 Aug 2011 13:38:07 +0200
> From: Stefan Klein <s.klein at erasmusmc.nl>
> Subject: Re: [elastix] Elastix Ubuntu New User
> To: "Robinson, Jamie" <j.robinson.10 at aberdeen.ac.uk>
> Cc: "elastix at bigr.nl" <elastix at bigr.nl>
> Message-ID: <4E426D9F.5000704 at erasmusmc.nl>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Hi Jamie,
> That probably explains the error. You can better split the image stacks,
> and register the anterior views separately and the posterior views
> separately.
>
> Right now, I don't think elastix reads the images correctly even, since
> you use a 2d parameter file, and give a multi-image stack as input
> (which could be considered as a 3d image probably). Not sure what
> happens then.
>
> Hope this solves the problem.
> kind regards,
> Stefan
>
> On 10/08/2011 10:20, Robinson, Jamie wrote:
> > Hi Stefan,
> >
> > Both image are 256x1024 with voxel size 2.4x2.4x1 mm. Each file is an
> > image stack with 2 slices - an anterior and posterior view.
> >
> > Thanks,
> > Jamie
> >
> >
> > ------------------------------------------------------------------------
> > *From:* elastix-bounces at bigr.nl [elastix-bounces at bigr.nl] on behalf of
> > Stefan Klein [s.klein at erasmusmc.nl]
> > *Sent:* Tuesday, August 09, 2011 9:57 PM
> > *To:* elastix at bigr.nl
> > *Subject:* Re: [elastix] Elastix Ubuntu New User
> >
> > Hi Jamie,
> > Thanks for your interest in elastix. The warnings are not a problem.
> > They just inform you that some parameters were not specified and that
> > elastix takes a default value.
> > The error means that the fixed and moving image do not have a
> > sufficiently large initial overlap. Or maybe your moving image is much
> > smaller than your fixed image. Could you email us:
> > - The size of the images
> > - The voxel size
> > Kind regards,
> > Stefan
> >
> >
> > On 9-8-2011 19:55, Robinson, Jamie wrote:
> >> I am working on developing some software that compares two nuclear
> >> medicine whole body bone scans both intra and intersubject.
> >> Previously, I have used a registration package bUnwarpJ in ImageJ for
> >> this purpose. However, I was not satisfied with the results (it uses a
> >> simple sum of squared intensity differences as a metric) and would now
> >> like to try something a bit more sophisticated and stumbled across
> >> elastix.
> >>
> >> I followed the user manual on how to install the package and
> >> downloaded the default parameter configuration files. However, upon
> >> running the command
> >>
> >> elastix -f FS1.dcm -m FS2.dcm -p ParameterFiles/Parameters_Affine.txt
> >> -p ParameterFiles/Parameters_BSpline.txt -out .
> >>
> >> I am met with a whole bunch of warnings before the software fails. I
> >> have attached the log file but have no idea why the software fails.
> >> Can anyone offer any help.
> >>
> >> Thanks,
> >> Jamie
> >>
> >>
> >> _______________________________________________
> >> elastix mailing list
> >> elastix at bigr.nl
> >> http://lists.bigr.nl/mailman/listinfo/elastix
>
> --
> Stefan Klein
> +31 10 7043049
> http://www.bigr.nl/people/StefanKlein
>
>
> ------------------------------
>
> Message: 2
> Date: Wed, 10 Aug 2011 11:55:18 +0000
> From: "Robinson, Jamie" <j.robinson.10 at aberdeen.ac.uk>
> Subject: Re: [elastix] Elastix Ubuntu New User
> To: Stefan Klein <s.klein at erasmusmc.nl>
> Cc: "elastix at bigr.nl" <elastix at bigr.nl>
> Message-ID:
>        <
> AB45752CD13C724FB1AAA26FDE4F6E37019926EE at AMSPRD0104MB121.eurprd01.prod.exchangelabs.com
> >
>
> Content-Type: text/plain; charset="us-ascii"
>
> Hi Stefan,
>
> The first solution works perfectly. In fact I only need to register one of
> the views as both anterior and posterior views are acquired simultaneously.
> Therefore, once I find the deformation field that registers the anterior
> views I can quickly register the posterior views with the mirror
> transformation. I look forward to using Elastix to help me with my task.
>
> Thanks,
> Jamie
> ________________________________________
> From: Stefan Klein [s.klein at erasmusmc.nl]
> Sent: Wednesday, August 10, 2011 12:38 PM
> To: Robinson, Jamie
> Cc: elastix at bigr.nl
> Subject: Re: [elastix] Elastix Ubuntu New User
>
> Hi Jamie,
> That probably explains the error. You can better split the image stacks,
> and register the anterior views separately and the posterior views
> separately.
>
> Right now, I don't think elastix reads the images correctly even, since
> you use a 2d parameter file, and give a multi-image stack as input
> (which could be considered as a 3d image probably). Not sure what
> happens then.
>
> Hope this solves the problem.
> kind regards,
> Stefan
>
> On 10/08/2011 10:20, Robinson, Jamie wrote:
> > Hi Stefan,
> >
> > Both image are 256x1024 with voxel size 2.4x2.4x1 mm. Each file is an
> > image stack with 2 slices - an anterior and posterior view.
> >
> > Thanks,
> > Jamie
> >
> >
> > ------------------------------------------------------------------------
> > *From:* elastix-bounces at bigr.nl [elastix-bounces at bigr.nl] on behalf of
> > Stefan Klein [s.klein at erasmusmc.nl]
> > *Sent:* Tuesday, August 09, 2011 9:57 PM
> > *To:* elastix at bigr.nl
> > *Subject:* Re: [elastix] Elastix Ubuntu New User
> >
> > Hi Jamie,
> > Thanks for your interest in elastix. The warnings are not a problem.
> > They just inform you that some parameters were not specified and that
> > elastix takes a default value.
> > The error means that the fixed and moving image do not have a
> > sufficiently large initial overlap. Or maybe your moving image is much
> > smaller than your fixed image. Could you email us:
> > - The size of the images
> > - The voxel size
> > Kind regards,
> > Stefan
> >
> >
> > On 9-8-2011 19:55, Robinson, Jamie wrote:
> >> I am working on developing some software that compares two nuclear
> >> medicine whole body bone scans both intra and intersubject.
> >> Previously, I have used a registration package bUnwarpJ in ImageJ for
> >> this purpose. However, I was not satisfied with the results (it uses a
> >> simple sum of squared intensity differences as a metric) and would now
> >> like to try something a bit more sophisticated and stumbled across
> >> elastix.
> >>
> >> I followed the user manual on how to install the package and
> >> downloaded the default parameter configuration files. However, upon
> >> running the command
> >>
> >> elastix -f FS1.dcm -m FS2.dcm -p ParameterFiles/Parameters_Affine.txt
> >> -p ParameterFiles/Parameters_BSpline.txt -out .
> >>
> >> I am met with a whole bunch of warnings before the software fails. I
> >> have attached the log file but have no idea why the software fails.
> >> Can anyone offer any help.
> >>
> >> Thanks,
> >> Jamie
> >>
> >>
> >> _______________________________________________
> >> elastix mailing list
> >> elastix at bigr.nl
> >> http://lists.bigr.nl/mailman/listinfo/elastix
>
> --
> Stefan Klein
> +31 10 7043049
> http://www.bigr.nl/people/StefanKlein
>
>
> ------------------------------
>
> Message: 3
> Date: Wed, 10 Aug 2011 10:29:30 -0400
> From: Kristen Tgavalekos <kristen.tgav at gmail.com>
> Subject: [elastix] 3D Bspline registration
> To: elastix at bigr.nl
> Message-ID:
>        <CAH1hsYuG+BhYE-T0zom8ZFbJVhb1wBmWm0AMpovMsctRKait-g at mail.gmail.com
> >
> Content-Type: text/plain; charset="iso-8859-1"
>
> Hi Elastix team,
>
> I'm working on registering 3D images with a BSpline transform. They are
> limited angle tomography images with different depths (a 10 slice
> difference) and the resolution is 0.1 x 0.1 x 1 mm. My problem is that the
> final registered image has blank holes in it. Typically the top and bottom
> slices of the volume have the most holes. When I use the same parameters
> for
> 2D registration with slices from these images, the registered image is well
> aligned with the static image and there are no holes. Does any one know
> what
> could be the reason for the errors I'm getting with 3D registration? As a
> side note, I've had a similiar issue with affine transforms but instead of
> holes, the slices in the final image look as if they were cropped.
>
> Here's some more info about my parameters if it helps:
> Optimizer: advanced stochastic gradient descent with automatic parameter
> estimation
> Metric: advanced mattes mutual information
> Image sampler: random coordinate
> Number of iterations: 1000
> Number of resolutions: 10
>
> Thank you,
> Kristen
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> http://lists.bigr.nl/pipermail/elastix/attachments/20110810/14f21a66/attachment.html
>
> ------------------------------
>
> Message: 4
> Date: Wed, 10 Aug 2011 22:36:08 +0200
> From: Stefan Klein <s.klein at erasmusmc.nl>
> Subject: Re: [elastix] 3D Bspline registration
> To: elastix at bigr.nl
> Message-ID: <4E42EBB8.7000106 at erasmusmc.nl>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Hi Kristen,
>
> This is probably because of motion in the z-direction. At the boundary
> of the image, the transformation is likely to map outside the moving
> image. In the result image this looks like a hole. To verify that this
> is the case, you could set the (DefaultValue ... ) entry in your
> parameter file to some specific value, for example 1000, and check if
> you get these values in the holes. The default value is the value that
> elastix uses for pixels that map outside the moving image domain (when
> generating the deformed moving image).
>
> If you think that motion out of plane is not appropriate in your
> application (due to the thick slice thickness for example), you could
> add the following line to your parameter file:
> (MovingImageDerivativeScales 1 1 0)
>
> Then elastix will not make any displacement in the z-direction.
>
> Kind regards,
> Stefan
>
> On 10-8-2011 16:29, Kristen Tgavalekos wrote:
> > Hi Elastix team,
> >
> > I'm working on registering 3D images with a BSpline transform. They
> > are limited angle tomography images with different depths (a 10 slice
> > difference) and the resolution is 0.1 x 0.1 x 1 mm. My problem is that
> > the final registered image has blank holes in it. Typically the top
> > and bottom slices of the volume have the most holes. When I use the
> > same parameters for 2D registration with slices from these images, the
> > registered image is well aligned with the static image and there are
> > no holes. Does any one know what could be the reason for the errors
> > I'm getting with 3D registration? As a side note, I've had a similiar
> > issue with affine transforms but instead of holes, the slices in the
> > final image look as if they were cropped.
> >
> > Here's some more info about my parameters if it helps:
> > Optimizer: advanced stochastic gradient descent with automatic
> > parameter estimation
> > Metric: advanced mattes mutual information
> > Image sampler: random coordinate
> > Number of iterations: 1000
> > Number of resolutions: 10
> >
> > Thank you,
> > Kristen
> >
> >
> > _______________________________________________
> > elastix mailing list
> > elastix at bigr.nl
> > http://lists.bigr.nl/mailman/listinfo/elastix
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> http://lists.bigr.nl/pipermail/elastix/attachments/20110810/d3a81333/attachment-0001.htm
>
> ------------------------------
>
> _______________________________________________
> elastix mailing list
> elastix at bigr.nl
> http://lists.bigr.nl/mailman/listinfo/elastix
>
>
> End of elastix Digest, Vol 26, Issue 5
> **************************************
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bigr.nl/pipermail/elastix/attachments/20110815/5714ddce/attachment.html>


More information about the Elastix mailing list