antsJointTensorFusion(1) part of ANTS registration suite



antsJointTensorFusion is an image fusion algorithm developed by Hongzhi Wang and Paul Yushkevich which won segmentation challenges at MICCAI 2012 and MICCAI 2013. The original label fusion framework was extended to accommodate intensities by Brian Avants. This implementation is based on Paul's original ITK-style implementation and Brian's ANTsR implementation. References include 1) H. Wang, J. W. Suh, S. Das, J. Pluta, C. Craige, P. Yushkevich, Multi-atlas segmentation with joint label fusion IEEE Trans. on Pattern Analysis and Machine Intelligence, 35(3), 611-623, 2013. and 2) H. Wang and P. A. Yushkevich, Multi-atlas segmentation with joint label fusion and corrective learning--an open source implementation, Front. Neuroinform., 2013.


-d, --image-dimensionality 2/3/4
This option forces the image to be treated as a specified-dimensional image. If not specified, the program tries to infer the dimensionality from the input image.
-t, --target-image targetImage
The target image (or multimodal target images) assumed to be aligned to a common image domain.
-g, --atlas-image atlasImage
The atlas image (or multimodal atlas images) assumed to be aligned to a common image domain.
-l, --atlas-segmentation atlasSegmentation
The atlas segmentation images. For performing label fusion the number of specified segmentations should be identical to the number of atlas image sets.
-a, --alpha 0.1
Regularization term added to matrix Mx for calculating the inverse. Default = 0.1
-b, --beta 2.0
Exponent for mapping intensity difference to the joint error. Default = 2.0
-r, --retain-label-posterior-images (0)/1
Retain label posterior probability images. Requires atlas segmentations to be specified. Default = false
-f, --retain-atlas-voting-images (0)/1
Retain atlas voting images. Default = false
-c, --constrain-nonnegative (0)/1
Constrain solution to non-negative weights.
-u, --log-euclidean (0)/1
Use log Euclidean space for tensor math
-p, --patch-radius 2
Patch radius for similarity measures. Default = 2x2x2
-m, --patch-metric (PC)/MSQ
Metric to be used in determining the most similar neighborhood patch. Options include Pearson's correlation (PC) and mean squares (MSQ). Default = PC (Pearson correlation).
-s, --search-radius 3
Search radius for similarity measures. Default = 3x3x3
-e, --exclusion-image label[exclusionImage]
Specify an exclusion region for the given label.
-x, --mask-image maskImageFilename
If a mask image is specified, fusion is only performed in the mask region.
-o, --output labelFusionImage
intensityFusionImageFileNameFormat [labelFusionImage,intensityFusionImageFileNameFormat,<labelPosteriorProbabilityImageFileNameFormat>,<atlasVotingWeightImageFileNameFormat>]
The output is the intensity and/or label fusion image. Additional optional outputs include the label posterior probability images and the atlas voting weight images.
Get version information.
-v, --verbose (0)/1
Verbose output.
Print the help menu (short version).
Print the help menu. <VALUES>: 1