[elastix] A question

M.Staring at lumc.nl M.Staring at lumc.nl
Thu Mar 18 09:49:40 CET 2010


Hi Andriy,

> 
> Hi Marius,
> 
> You've been advising me about "local MI" Elastix 
> implementation previously. Thank you. I have a few general 
> questions through.
> 
> As far as I understand the way the local MI is implemented is 
> following:
> At each iteration Elastix select a random voxel, then sample, 
> e.g. 2000 samples randomly from the square neighborhood (e.g 
> 50x50x50) around that voxel, and use this information for 
> optimization (using StandardGradientDescend).
> Am I correct here?

That is correct!

You can find the precise implementation in the files

	
elastix\src\Common\ImageSamplers\itkImageRandomCoordinateSampler.h
	
elastix\src\Common\ImageSamplers\itkImageRandomCoordinateSampler.txx

The RandomSampleRegion option seems to be only implemented for the
RandomCoordinate sampler.

Note that you can use more optimizers than just StandardGradientDescend
with this specific way of sampling.

> 
> Then my questions are:
> 
> 1) Can I use not just 1 single voxel (and its neighborhood) 
> per iterations, but use a sum of e.g. 100 of neighborhoods 
> simultaneously?
> or even all the image voxels (and its neighborhoods)?

Nope. Currently, this sampler simply samples from one (1) region.

You can however easily create a new or modify the afforementioned
sampler, which implements your wishes. It only requires modification of
the GenerateData() function of this class. And some changes in 

	elastix\src\Components\ImageSamplers\RandomCoordinate

We can give you some pointers if you are interested.

> 
> 2) When the random voxel and its neighborhood is selected can 
> I use simply all the voxel within?

Also a nope. And again it is quite easy to change the code.

Stefan may be able to give some more comments, since he is the one that
implemented locality in the image samplers.

Hope this helps,

Marius

> 
> Thank you for the answer. I really appreciate the code you've 
> made publicly available. The questions I've asked are not to 
> criticize anything, it's just to clarify the Elastix 
> implementation for me personally.
> 
> ---
> Sincerely,
> Andriy Myronenko
> Oregon Health and Science University
> www.bme.ogi.edu/~myron
> 
> 
> 
> 
> M.Staring at lumc.nl wrote:
> > Hi Andriy,
> > 
> > Yes this is indeed included in the latest release!
> > 
> > This page will give you some information about the 
> available parameters:
> > 
> > 	http://elastix.isi.uu.nl/doxygen/a00053.html#_details
> > 
> > And at the parameter file database of elastix I have uploaded the 
> > exact configurations that I used for that paper:
> > 
> > 	http://elastix.bigr.nl/wiki/index.php/Par0005
> > 
> > We have some good experience with elastix with local similarity 
> > metrics (e.g. local mutual information) to compensate for 
> > non-homogeneity in MR images. See this paper:
> > 
> > S. Klein, U.A. van der Heide, I.M. Lips, M. van Vulpen, M. 
> Staring and 
> > J.P.W. Pluim, "Automatic Segmentation of the Prostate in 3D 
> MR Images 
> > by Atlas Matching using Localised Mutual Information," Medical 
> > Physics, vol. 35, no. 4, pp. 1407 - 1417, April 2008.
> > 
> > Local metrics can be selected in elastix by setting the following 
> > sampling parameters in the parameter text file:
> > 
> > (UseRandomSampleRegion "true")
> > (SampleRegionSize 50.0 50.0 50.0)
> > 
> > where UseRandomSampleRegion sets locality to on, and where 
> > SampleRegionSize sets the size of the region on which a local 
> > similarity measure is computed (in physical units, usually mm).
> > 
> > If you need any help, please join our mailing list and post your 
> > questions :-)
> > 
> > 	http://lists.bigr.nl/mailman/listinfo/elastix
> > 
> > With kind regards,
> > 
> > Marius Staring
> > 
> > 
> >> -----Original Message-----
> >> From: Andriy Myronenko [mailto:myron at csee.ogi.edu]
> >> Sent: donderdag 22 oktober 2009 2:52
> >> To: Staring, M. (LKEB)
> >> Subject: A question
> >>
> >> Dear Marius Staring,
> >>
> >> Good time of the day. I have a question regarding the Elastix 
> >> software for image registration.
> >>
> >> Does it include the code related to the recent paper:
> >> "Registration of Cervical MRI Using Multifeature Mutual 
> Information" ?
> >>
> >> If not, does Elastix include any similarity metric to deal with 
> >> intensity non-homogeneity (bias distortions)?
> >>
> >> Thanks for you reply
> >>
> >> --
> >> Sincerely,
> >> Andriy Myronenko
> >> Oregon Health and Science University
> >>
> >> No virus found in this incoming message.
> >> Checked by AVG - www.avg.com
> >> Version: 8.5.423 / Virus Database: 270.14.25/2450 - Release
> >> Date: 10/21/09 16:44:00
> >>
> 
> --
> Sincerely,
> Andriy Myronenko
> 



More information about the Elastix mailing list