[Insight-users] Scaling in Registration

[Michael Hardisty]

Hello Luis and others,

Thanks for your quick and helpful response.

Answer: 1) When I say smaller, I mean physically (measured in 
millimeters) smaller.  The image happens to also have less pixels 
because the region I am concerned with is physically smaller and the 
pixel spacing remains the same because of the uCT detector.  The reason 
that I mentioned pixel densities is that if I manually scaled my images 
to align them the pixel spacings of the two images would differ and 
hence would have different pixel spacings.  I previously used the word 
densities as a synonym for spacing I apologize for the confusion.

I am currently working with a downsampled version of the images that I 
will eventually have to deformably register.  I am also working with 
Images that I have artificially deformed to make a set of images that I 
believe will be similar to the ones I will eventually obtain.  Please 
note that I am most interested in the mapping between the two images and 
am less interested in what the images actually look like.  I am looking 
at the images simply to verify the mapping.


Test Case Images(downsampled):

Moving Image:                                         Target Image:
                                XxYxZ                                   
                  XxYxZ
Pixel Number(#):      50x50x50                 Pixel Number(#):      
55x55x55
Pixel Spacing(mm):   0.19x0.19x0.19        Pixel Spacing(mm):   
0.19x0.19x0.19
Origin(mm):              17,10,14                   Origin(mm):         
     16.6,9.7,13.6

Real Images(not downsampled):

Moving Image:                                         Targest Image:
                                XxYxZ                                   
                  XxYxZ
Pixel Number(#):      260x280x280                 Pixel Number(#):      
286x310x310
Pixel Spacing(mm):   0.035x0.035x0.035        Pixel Spacing(mm):   
0.035x0.035x0.035
Origin(mm):              17,10,14                   Origin(mm):         
     16.5,9.5,13.5

Thanks for your help with this matter.

-- 
Michael Hardisty
M.A.Sc Student
University of Toronto
Orthopaedic Biomechanics Laboratory
Sunnybrook & Women's College Health Sciences Centre



Luis Ibanez wrote:

>
>
> Hi Michael,
>
>
>
> When you say "smaller" do you mean:
>
>
> 1) The physical extent of the image measured in
>    millimeters is smaller
>
>
>                 OR do you mean:
>
>
> 2) The number of pixels of the image is smaller.
>
>
>
> You description lead us to think that you are registering
> two images that have different pixel spacing and you are
> trying to interpret the scaling in terms of pixels instead
> of in terms of millimeters (in the physical world).
>
>
> Please post to the list the following information for both
> the Fixed and Moving images:
>
>
>     1) Number of pixels along each dimension
>     2) Pixel spacing in millimeters
>     3) Origin of the image in millimeters
>
>
>
>
> You *MUST* look at the following section of our course on
> image registration:
>
> http://www.cs.rpi.edu/courses/spring04/imagereg/lecture07.ppt
>
> and you *MUST* read the following sections
> of the ITK Software Guide:
>
>
>       http://www.itk.org/ItkSoftwareGuide.pdf
>
>
>   A) Section 6.7.1, "Geometric Transformations"  pdf-pages 199-219
>   B) Chapter 8. "Registration", pdf-pages 241-340
>
>
>
> Regards,
>
>
>
>    Luis
>
>
>
>
> -----------------------------
> Michael Hardisty wrote:
>
>> I am attempting to use a centred affine transform to register two 3D 
>> volumes.  I believe that the registration should be a combination of 
>> shearing, scaling, rotation and translation.  Therefore the Affine 
>> transform seems like the ideal choice.  The problem I am having is 
>> that when I do the registration I never get proper scaling of the 
>> volumes.  The Image I am registering is smaller than the target image 
>> and remains smaller after the registration has converged.  The other 
>> parts of the transformation look relatively good, the scaling is the 
>> only part of the registration that seems totally off.  I thought that 
>> the scaling might be suppressed by the particular metric that I am 
>> using (MeanSquaresImageToImageMetric), but I am not sure.  This leads 
>> me to several questions:
>>
>> 1)  How do the metrics evaluate when the images have different pixel 
>> densities?  Does it make sense to suggest that by stretching the 
>> moving image over more target pixels that this could cause a greater 
>> metric value because more pixels are included?  If so which metric is 
>> the most suited to deal with this problem?
>>
>> 2)  Are there any other parts of the registration process that may 
>> effect the scaling of the moving image?  What other strategies could 
>> I take to encourage more scaling?
>>
>> Current Registration Components:
>> Metric:         MeanSquaresImageToImageMetric
>> Interpolator:  LinearInterpolateImageFunction
>> Optimizer:     RegularStepGradientDescentOptimizer
>> Transform:    CenteredAffineTransform
>> Image Types: uCT (pixel type = float)
>>
>> Thanks
>>
>
>
>
>

-- 
Michael Hardisty
M.A.Sc Student
University of Toronto
Orthopaedic Biomechanics Laboratory
Sunnybrook & Women's College Health Sciences Centre