[Insight-users] 3D image registration with masking

[Angela Lee]

Hi all,

I am trying to rigidly register 3D (gipl) images using a mask, I have
used Normalised Correlation (NC) and Normalised Mutual Information (NMI)
metrics by modifying ImageRegistration14.cxx and ImageRegistration8.cxx,
I have attached my versions of the code.
I am using NMI as I would like to extend into other imaging modalities
later on.

I tried out the code on the same image to make sure I get the ideal
value of the metric (-1 for NC and 2 for NMI)

My code works fine when I use a very simple mask, that has 50% white
pixels.
I created a 3D mask using a stack of 2D images I generated using another
software program.
However when I try to use my 3D mask I get this error for my code using
the NMI metric:

------------------------
Initial transform parameters = [0, 0, 0, 0, 0, 0]
optimizerScales = [10000, 10000, 10000, 0.0599532, 0.0731429, 0.0769231]
ExceptionObject caught !

itk::ExceptionObject (0x85cf650)
Location: "void itk::HistogramImageToImageMetric<TFixedImage,
TMovingImage>::ComputeHistogram(const typename
itk::ImageToImageMetric<TFixedImage,
TMovingImage>::TransformParametersType&,
itk::Statistics::Histogram<double, 2u,
itk::Statistics::DenseFrequencyContainer>&) const [with TFixedImage =
itk::Image<unsigned char, 3u>, TMovingImage = itk::Image<unsigned char,
3u>]"
File:
/people/alee071/Desktop/itk/InsightToolkit-3.2.0/Code/Algorithms/itkHistogramImageToImageMetric.txx
Line: 303
Description: itk::ERROR:
NormalizedMutualInformationHistogramImageToImageMetric(0x85c13b8): All
the points mapped to outside of the moving image
------------------------

I also have an issue with using my normalised correlation metric code
with the mask, when I try without a mask, I get :
Metric value (ideal = -1) = -1

But when I try with my mask I get :
Metric value (ideal = -1) = 0

I think my problem may be that my 3D mask has a low percentage of white
pixels.

My code and the images can be found on:
ftp://ftp.bioeng.auckland.ac.nz/alee071/

Is there anyway I can get past this?
Any help is greatly appreciated. :)

Thanks,
Angela