Output image similarity metrics

[nadluru]

Dear everyone,

Hope everyone is doing well.

I was wondering if there are command line tools within the ANTS framework to output image similarity measures between two images such as normalized mutual information. I noticed ImageMath has Mattes operation, but it doesn't seem to provide an option to normalize by the sum of the entropies of the two images. I also noticed that the NormalizedCorrelation operation in ImageMath outputs -1 for two identical inputs for which the value would be expected to be 1.

It would be wonderful to have a command like

comparison_command 3 output.txt similarity_metric image1 image2 {optional_mask}

where similarity_metric could be any of the options allowed for "--metric" in antsRegistration.

Could you kindly let me know your thoughts, questions, and suggestions? Thank you so much for your guidance and help.

Sincerely,
Nagesh

[ntustison]

% MeasureImageSimilarity 

COMMAND: 
     MeasureImageSimilarity
          Program to calculate the similarity between two images using various metrics. 

OPTIONS: 
     -d, --dimensionality 2/3/4
          Dimensionality of the fixed/moving image pair. 

     -m, --metric CC[fixedImage,movingImage,metricWeight,radius,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  MI[fixedImage,movingImage,metricWeight,numberOfBins,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  Mattes[fixedImage,movingImage,metricWeight,numberOfBins,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  MeanSquares[fixedImage,movingImage,metricWeight,radius=NA,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  Demons[fixedImage,movingImage,metricWeight,radius=NA,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  GC[fixedImage,movingImage,metricWeight,radius=NA,<samplingStrategy={None,Regular,Random}>,<samplingPercentage=[0,1]>]
                  ICP[fixedPointSet,movingPointSet,metricWeight]
                  PSE[fixedPointSet,movingPointSet,metricWeight,<pointSetSigma=1>,<kNeighborhood=50>]
                  JHCT[fixedPointSet,movingPointSet,metricWeight,<pointSetSigma=1>,<kNeighborhood=50>,<alpha=1.1>,<useAnisotropicCovariances=1>]
          These image metrics are available--- CC: ANTS neighborhood cross correlation, 
          MI: Mutual information, Demons: (Thirion), MeanSquares, and GC: Global 
          Correlation. The "metricWeight" variable is not used. The metrics can also 
          employ a sampling strategy defined by a sampling percentage. The sampling 
          strategy defaults to 'None' (aka a dense sampling of one sample per voxel), 
          otherwise it defines a point set over which to optimize the metric. In addition, 
          three point set metrics are available: Euclidean (ICP), Point-set expectation 
          (PSE), and Jensen-Havrda-Charvet-Tsallis (JHCT). 

     -x, --masks fixedImageMask
                 [fixedImageMask,movingImageMask]
                 [fixedImageMask,NULL]
          Image masks to limit voxels considered by the metric. 

     -v, --verbose (0)/1
          Verbose output. 

     -o, --output gradientImage
          Output the metric gradient image (optional). 

     -h 
          Print the help menu (short version). 

     --help 
          Print the help menu. 

[nadluru]

Oh wow! I didn't know about this command. I am so excited to try this. Thank you so much, Nick. Really appreciate the prompt help!

Sincerely,
Nagesh

[nadluru]

Hi Nick,

So sorry to bother you again. The MeasureImageSimilarity command is working nicely. Bot I was wondering if there is an option for normalized mutual information where the values get scaled to be between 0 and 1. My understanding is that we could get it by using the following formula.

NMI = 2 * MI / (entropy of image 1 + entropy of image 2)

Is there an ANTs command to compute entropy of a given image?

Thanks again so much!

Sincerely,
Nagesh

[ntustison]

ImageIntensityStatistics

[nadluru]

Oh wow! This is simply amazing, Nick. Thank you so much again!

Sincerely,
Nagesh