[Insight-users] [Insight-registration] Registration of one sub-brain to full-brain

[brian avants]

Hi Torsten

We've leaned towards the ability to provide explicit mask(s) to the
registration and metrics that allow you to focus the registration on
regions of interest.  It's an easy extension with potentially low
overhead to computation time, even in the dense case ... we should be
able to set up the image iterators to run only through the ROI and/or
overlap of ROIs, the latter meaning the overlap of the two masks you
passed in after transformation to a common space ... the latter part
is a little more interesting to think about in that the overlap of
ROIs will change at every iteration.   I think the optimized image
registration metrics in v4 allow something like this already.

B

On Thu, Oct 21, 2010 at 1:17 PM, Torsten Rohlfing
<torsten at synapse.sri.com> wrote:
> Hi guys -
>
> This is more of a question than a suggestion: does ITK currently support
> "pixel padding"? That is, can we currently define a value in the image data
> that serves as a marker for pixels that aren't really there? Think of it as
> a binary mask coded into the image itself.
>
> In my experience, when you register partial objects to complete objects
> (also interesting to register skull-stripped brains to full head images, for
> example), the key is to be able to exclude pixels from the metric
> computation altogether, rather than just setting them to zero.
>
> Alternatively, because we cannot always find a pixel value that isn't
> actually used by the data already, one could implement this behaviour by
> providing the registration framework with optional explicit binary masks for
> fixed and moving images.
>
> Best,
>   Torsten
>
> Hello Erik
>
> You may want to try changing the number of resolutions employed in the
> multi-resolution pyramid used to perform the registration.   In
> addition, you may want to try switching the role of the fixed/moving
> image.  We are currently revising the registration framework so now is
> a good time to bring these types of issues up and work with us on
> improving registration performance.
>
> Brian
>
>
> On Thu, Oct 21, 2010 at 4:35 AM, Erik Tuerke <tuerke at cbs.mpg.de> wrote:
>> Hi!
>>
>> I want to register one hemisphere (actually 30 slices of 0.75 mm) to a
>> fullbrain scan (30 slices of 0.75mm, but shifted to superior for about 5
>> mm)
>> . Currently i am using a MattesMutualInformationImageToImage metric, a
>> RegularStepGradientDescentOptimizer and a VersorRigid3DTransform.
>>
>> Prior to the registration the hemisphere is roughly aligned to the
>> fullbrain
>> (only needs translation of few mm and little rotation). If i start the
>> registration the optimizer finishes with a rotation of about 90 deg x and
>> y.
>> I also tried a VersorRigid3DOptimizer and a NormalizedCorrelation with
>> almost the same result.
>>
>> The scalefactors of my optimizer are:
>>               optimizerScaleRegularStepGradient[0] = 1.0;
>>               optimizerScaleRegularStepGradient[1] = 1.0;
>>               optimizerScaleRegularStepGradient[2] = 1.0;
>>               optimizerScaleRegularStepGradient[3] = 1.0/1000.0;
>>               optimizerScaleRegularStepGradient[4] = 1.0/1000.0;
>>               optimizerScaleRegularStepGradient[5] = 1.0/1000.0;
>>
>> I guess the itk examples are proposing the same values.
>> Is there any trick to apply such a sub-brain registration to a full-brain
>> registration. The worrying thing is that SPM gets it right :-(
>>
>> Thanks and best regards!!
>>
>> --
>> Erik Tuerke
>> Department of Neurophysics Max-Planck-Institute for Human Cognitive and
>> Brain Sciences
>> Stephanstrasse 1A
>> 04103 Leipzig
>> Germany Tel: +49 341 99 40-2440
>> Email: tuerke at cbs.mpg.de
>> www.cbs.mpg.de
>>
>>
>> _____________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Kitware offers ITK Training Courses, for more information visit:
>> http://www.kitware.com/products/protraining.html
>>
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-users
>>
>
>
>
> --
> Torsten Rohlfing, PhD          SRI International, Neuroscience Program
>  Senior Research Scientist      333 Ravenswood Ave, Menlo Park, CA 94025
>   Phone: ++1 (650) 859-3379      Fax: ++1 (650) 859-2743
>    torsten at synapse.sri.com        http://www.stanford.edu/~rohlfing/
>
>      "Though this be madness, yet there is a method in't"
>
> _______________________________________________
> Insight-registration mailing list
> Insight-registration at public.kitware.com
> http://public.kitware.com/cgi-bin/mailman/listinfo/insight-registration
>
>



-- 
ß®∫∆π