[Insight-users] Difference between 3.4 and 3.14 for MattesMutualInformation Registration ?

[Hans Johnson]

Uma,

In 3.14 there were several bug fixes that allow the methods to work with
oriented images.  

If you have the opportunity, I¹d recommend the following CMakeFlags to get
the best performance, and the proper interpretation of image physical space
for oriented images:

                          -DITK_USE_OPTIMIZED_REGISTRATION_METHODS:BOOL=ON \
                          -DITK_USE_REVIEW:BOOL=ON \
                          -DITK_USE_TRANSFORM_IO_FACTORIES:BOOL=ON \
                          -DITK_USE_ORIENTED_IMAGE_DIRECTION:BOOL=ON \
                          -DITK_IMAGE_BEHAVES_AS_ORIENTED_IMAGE:BOOL=ON \
                          -DITK_USE_PORTABLE_ROUND:BOOL=ON \
                   
-DITK_USE_CENTERED_PIXEL_COORDINATES_CONSISTENTLY:BOOL=ON \

You may want to look at the project that we¹ve used extensively to do this
for brain images:
http://www.nitrc.org/projects/multimodereg/

In creating the BRAINSFit package, we developed an extensive test suite to
ensure that the registrations are occurring properly for a wide class of
image physical space orientations.

Hans



On 3/28/10 6:57 AM, "Uma Ranjan" <usranjan at gmail.com> wrote:

> Hi Luis,
>  
>  No, the images have (1,0,0/0,1,0/0,0,1) as direction cosines. Also, the
> images are from an in-line system which has identical slice spacing, so they
> are more or less registered already. I found that the 1.3 version of the same
> optimizer gave much better results.
>  
> And no - I have not turned ON REVIEW or
> ITK_USE_OPTIMIZED_REGISTRATION_METHODS. Will they help ? I only found that I
> was not setting the direction in the resampler, but setting it only made
> things worse for the rotation values.
>  
> regards
> uma
> 
> On Sat, Mar 27, 2010 at 11:32 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>> Hi Uma,
>> 
>> Do your images have Orientation ?
>> 
>> That is, the value of their Direction matrix,
>> is it different from Identity ?
>> 
>> If so, the behavior of ITK 3.14 will be more
>> appropriate than the behavior of ITK 3.4.
>> 
>> Also:
>> 
>> Have you turned ON the use of the REVIEW
>> directory ?
>> 
>> are you using the flag
>> 
>> ITK_USE_OPTIMIZED_REGISTRATION_METHODS  ?
>> 
>> 
>> ---
>> 
>> To assist you better, we will need more details
>> about the registration that you are performing.
>> 
>> 
>>      Thanks
>> 
>> 
>>             Luis
>> 
>> 
>> -----------------------------------------------------------------------
>> On Sat, Mar 27, 2010 at 1:26 PM, Uma Ranjan <usranjan at gmail.com> wrote:
>>> > Dear all,
>>> >
>>> >
>>> >  I see a difference in the transform values (and visual performance)
>>> between
>>> > ITK 3.4 and 3.14 versions for the same images and identical parameters.
>>> >
>>> >
>>> >
>>> >  I use VersorRigid3DTransform  with LinearInterpolateImageFunction as
>>> > interpolator,  MattesMutualInformationImageToImageMetric as metric and
>>> > VersorRigid3DTransformOptimizer as optimizer.
>>> >
>>> > The performance seemed to be better in 3.4. Is there anything additional
>>> > that I need to specify between the two versions ?
>>> >
>>> > I have set the following parameters for the optimizer :
>>> >    SetScales(optimizerScales), setMaximumStepLength(.),
>>> > setMinimumStepLength(.),SetNumberOfIterations(.), setRelaxationFactor(.),
>>> > MinimizeOn() ;
>>> >
>>> > For the Metric, I set the following :
>>> >
>>> >  SetNumberOfHistogramBins(), SetNumberOfSpatialSamples(.),
>>> > ReiniitalizeSeed(.),
>>> >
>>> > Best regards
>>> >
>>> > Uma Ranjan
>>> >
>>> > _____________________________________
>>> > Powered by www.kitware.com <http://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
>>> >
>>> >
> 
> 
> 
> _____________________________________
> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100328/82e7e31e/attachment.htm>