Is your feature request related to a problem? Please describe.
short question: Is there a way using ANTsR to edit the voxel dimensions of an image? Alternatively, is there a way to remove only the overall size component of a registration mapping (leaving the changes that remain after size has been equalized)?
longer explanation: I'm using ANTsR tools to register binary images (endocast) and measure differences between them in various ways (absolute distances, jacobians). These vary in size, and for some kinds of comparisons I want to focus on shape differences only vs. shape+size. A crude way to do this is to equalize the objects size (e.g., volume). I can do this by editing one of the image's voxel dimensions artificially, such that the volume of the binary object within it ends up being equal to the volume of the other object image. I found a way to do this using a command line tool from FSL (fsledithd) outside of ANTsR, but was wondering if there was a way to do this within ANTsR itself (so I can make a pipeline). Alternatively, is there a way to remove the size-only component of a registration mapping, i.e., after registering two images of different sizes, and just assess (from that mapping) the shape-only component? I know there is an affine and a diffeomorphic component to a combined mapping, but the affine part (as I understand it) could include, e.g., contraction/expansion along one dimension only. This would not be the same as changing volume (unless all dimensions of the affine transformation happen to be judged by the algorithm to be equivalent).
Describe the solution you'd like
1) An ANTsR function that lets you set the voxel dimensions to something new (maybe it already exists?), or
2) A way to constrain the affine mapping to be equivalent in all directions
Describe alternatives you've considered
Using FSL's fsledithd tool to make one of the image's voxel dimensions such that the binary object within it has the same volume as the binary object in the other image that I want to register together.
Is your feature request related to a problem? Please describe. short question: Is there a way using ANTsR to edit the voxel dimensions of an image? Alternatively, is there a way to remove only the overall size component of a registration mapping (leaving the changes that remain after size has been equalized)?
longer explanation: I'm using ANTsR tools to register binary images (endocast) and measure differences between them in various ways (absolute distances, jacobians). These vary in size, and for some kinds of comparisons I want to focus on shape differences only vs. shape+size. A crude way to do this is to equalize the objects size (e.g., volume). I can do this by editing one of the image's voxel dimensions artificially, such that the volume of the binary object within it ends up being equal to the volume of the other object image. I found a way to do this using a command line tool from FSL (fsledithd) outside of ANTsR, but was wondering if there was a way to do this within ANTsR itself (so I can make a pipeline). Alternatively, is there a way to remove the size-only component of a registration mapping, i.e., after registering two images of different sizes, and just assess (from that mapping) the shape-only component? I know there is an affine and a diffeomorphic component to a combined mapping, but the affine part (as I understand it) could include, e.g., contraction/expansion along one dimension only. This would not be the same as changing volume (unless all dimensions of the affine transformation happen to be judged by the algorithm to be equivalent).
Describe the solution you'd like 1) An ANTsR function that lets you set the voxel dimensions to something new (maybe it already exists?), or 2) A way to constrain the affine mapping to be equivalent in all directions
Describe alternatives you've considered Using FSL's fsledithd tool to make one of the image's voxel dimensions such that the binary object within it has the same volume as the binary object in the other image that I want to register together.
Thanks for any suggestions!