[Elastix] problem with non-rigid

Polfliet Mathias mpolflie at etro.vub.ac.be
Tue Feb 24 18:01:40 CET 2015


Dear Esmeralda,

It looks like your working on binary data. In that case I see a couple of issues with your parameter file. First of all your gridspacingschedule implies a first grid spacing of 16 voxels and then 8 voxels. Second, your multiresolutionscheme is not really multi resolution since your first resolution uses the original dimensions with interpolation and your second resolution uses the same dimension without interpolation. Which brings me to the final issue that you use a second order B-spline interpolation on a binary image?

With kind regards,

Mathias Polfliet

PhD Candidate on Medical Image Analysis
Department of Electronics and Informatics (ETRO), Vrije Universiteit Brussel
Office Ke.2.01, Pleinlaan 2, B-1050 Brussels, Belgium
T: +32 2 629 10 32 M: +32 475 48 66 14



On 24 Feb 2015, at 15:34, Floris Berendsen <floris at isi.uu.nl<mailto:floris at isi.uu.nl>> wrote:

Hi Esmeralda,

By default the registration procedure has no incentive to keep the borders of the images linked to each other.
If you want to keep all deformations at zero at the border, you could have a look at the setting PassiveEdgeWidth [1], but I never tried that.

[1] http://elastix.isi.uu.nl/doxygen/classelastix_1_1AdvancedBSplineTransform.html

Best,
Floris

On Tue, Feb 24, 2015 at 2:34 PM, Esmeralda Ruiz Pujadas <esmeralda.ruizpujadas at gmail.com<mailto:esmeralda.ruizpujadas at gmail.com>> wrote:
Dear Floris,

Thank you so much for your answer, so I suppose that it is located points away from the moving
domain so it is a way to say undefined. But I wonder why it does this, is there any way
to solve this problem?.

Thanks again !!
Sicnerely
Esmeralda

On Tue, Feb 24, 2015 at 2:26 PM, Floris Berendsen <floris at isi.uu.nl<mailto:floris at isi.uu.nl>> wrote:
Hi Esmeralda,

The black hole could be caused by the finite domain of the moving image. You can see if this is the case by setting the DefaultPixelValue to a specific value, e.g. 10. Any pixel with this value showing up in the result image is likely to be from outside the original moving domain and cannot be considered part of the object nor the background, but must be considered 'undefined'.

Best,
Floris

On Tue, Feb 24, 2015 at 12:57 PM, Esmeralda Ruiz Pujadas <esmeralda.ruizpujadas at gmail.com<mailto:esmeralda.ruizpujadas at gmail.com>> wrote:
Dear all,

I am writing because I got some artifacts in non-rigid registration
and I am wondering if someone had the same problem. I attach
the image the moving obtect is applying a non-rigid registration
using the script below, it is simply a non registration having
certain limits in their bending as i do not want to lose the details.
But as a results the first 3 frames I get that the biggest object
is cropped and only the border is presenved when the other moving image
is very similar to this one, so i do not understand this result.

Thank you so much
Sincerely
Esmeralda



// ********** Image Types

(FixedInternalImagePixelType "float")
(FixedImageDimension 3)
(MovingInternalImagePixelType "float")
(MovingImageDimension 3)
(UseDirectionCosines "true")
(Direction 1 0 0 0 1 0 0 0 1)
(ComputeZYX "false")

// ********** Components

(Registration "MultiMetricMultiResolutionRegistration")
(FixedImagePyramid "FixedSmoothingImagePyramid")
(MovingImagePyramid "MovingSmoothingImagePyramid")
(Transform "BSplineTransform")
(Metric "AdvancedMattesMutualInformation" "TransformBendingEnergyPenalty")
(Optimizer "AdaptiveStochasticGradientDescent")
(Interpolator "BSplineInterpolator")
(ResampleInterpolator "FinalBSplineInterpolator")
(Resampler "DefaultResampler")

// ********** Pyramid Scheme

//Identical scheme for fixed and moving image:
(NumberOfResolutions 2)
(ImagePyramidSchedule 1 1 1 0 0 0)

// ********** Transform

(BSplineTransformSplineOrder 3) //Default: 3
//Can also be specified with a parameter for each dimension
(FinalGridSpacingInVoxels 2.0)
(GridSpacingSchedule 8.0 4.0)
(HowToCombineTransforms "Compose")

// ********** Metric

//Can also be specified with a parameter for each resolution
(SubtractMean "true") //Default: true
(Metric0Weight 0.6)
(Metric1Weight 0.4)

// ********** Optimizer

//Can also be specified with a parameter for each resolution
(MaximumNumberOfIterations 250)
(AutomaticParameterEstimation "true")
(UseAdaptiveStepSizes "true")

(NumberOfJacobianMeasurements 65536)
(NumberOfSamplesForExactGradient 32768)

// ********** Masks

(ErodeMask "false")

// ********** ImageSampler

(ImageSampler "RandomCoordinate")
(NewSamplesEveryIteration "true")
(NumberOfSpatialSamples 32768)

(UseRandomSampleRegion "false")

// ********** Interpolator and Resampler

//B-Spline interp order used in each resolution:
//Can also be specified with a parameter for each resolution:
(BSplineInterpolationOrder 2)

//B-Spline interp order used for applying the final deformation:
(FinalBSplineInterpolationOrder 0)

//Default pixel value for pixels outside the picture:
(DefaultPixelValue 0)

// ********** Several/MISC

(WriteTransformParametersEachIteration "false")
(WriteTransformParametersEachResolution "false")
(WriteResultImageAfterEachResolution "false")
(WriteResultImage "true")
(CompressResultImage "false")
(ResultImageFormat "nii")
(ShowExactMetricValue "false")


_______________________________________________
Elastix mailing list
Elastix at bigr.nl<mailto:Elastix at bigr.nl>
http://lists.bigr.nl/cgi-bin/mailman/listinfo/elastix




_______________________________________________
Elastix mailing list
Elastix at bigr.nl<mailto:Elastix at bigr.nl>
http://lists.bigr.nl/cgi-bin/mailman/listinfo/elastix

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bigr.nl/pipermail/elastix/attachments/20150224/14a1c192/attachment.html>


More information about the Elastix mailing list