itk_similarity

calculate various ITK image similarity metrics between two volumes

itk_similarity <src> <target> [options]

DESCRIPTION

itk_similarity computes image similarity metrics between two input volumes using the Insight Toolkit (ITK). The tool supports a range of intensity-based and overlap-based metrics commonly used to evaluate registration quality, compare segmentation results, or measure inter-subject image correspondence.

Both volumes must be defined on the same sampling grid (same dimensions, voxel sizes, and orientation). An optional mask can restrict the computation to a specific region of interest.

Available metrics include mutual information (MI), normalized mutual information (NMI), Mattes mutual information (MMI), mean squared difference (MSQ), cross-correlation (CC), normalized cross-correlation (NCC), and Cohen’s kappa for label overlap (kappa).

OPTIONS

--verbose

Print verbose information during processing.

--clobber

Overwrite any output file if it already exists.

--mi

Compute mutual information (Viola-Wells formulation).

--nmi

Compute normalized mutual information.

--mmi

Compute Mattes mutual information.

--msq

Compute mean of squared differences.

--cc

Compute cross-correlation.

--ncc

Compute normalized cross-correlation.

--kappa

Compute Cohen’s kappa coefficient for label overlap comparison.

--mask file

Restrict the similarity computation to voxels within the given mask volume. Only voxels with non-zero values in the mask are included.

--bins n

Set the number of histogram bins used for mutual-information-based metrics. Default is 64.

EXAMPLES

Compute mutual information between two volumes:

itk_similarity source.mnc target.mnc --mi

Compute normalized cross-correlation within a brain mask:

itk_similarity source.mnc target.mnc --ncc --mask brain_mask.mnc

Compute Mattes mutual information with 128 histogram bins:

itk_similarity source.mnc target.mnc --mmi --bins 128

Compute kappa overlap for label volumes:

itk_similarity labels_a.mnc labels_b.mnc --kappa

Compute multiple metrics at once:

itk_similarity source.mnc target.mnc --mi --msq --ncc

AUTHOR

Vladimir S. Fonov - Montreal Neurological Institute, McGill University.

COPYRIGHTS

Copyright (C) Vladimir S. Fonov, McConnell Brain Imaging Centre, McGill University.

SEE ALSO

multiple_volume_similarity, minccalc, itk_resample, mincstats