Converting multi-component image to nii

[adityaapte]

I would like to write deformation vector field (dx,dy,dz volumes) as a multi-component image. Any suggestions on how to organize the input image and other metadata as an input to set_nii_hdr.m and nii_tool.m?

[xiangruili]

I am not clear about your purpose. Are you trying to store something like FSL warping file?

[adityaapte]

The requirement is to have 3 components per pixel - component 1: x-deformation. component 2: y-deformation and component 3: z-deformation. I was able to make progress by converting the DVF matrix to 8-D image as specified in help doc. The next issue is that the image pixel values get converted to uint8. Is there a setting to specify retaining the pixel intensities as float.
% Another way to signify RGB/RGBA data is to permute color dim to 8th-dim of img
% (RGB_dim of 8 can be omitted then). Since NIfTI img can have up to 7 dim,
% nii_tool chooses to store RGB/RGBA in 8th dim. Although this looks lengthy
% (4th to 7th dim are often all ones), nii_tool can deal with up to 7 dim
% without causing any confusion. This is why the returned nii.img always stores
% RGB in 8th dim.

[xiangruili]

I am not sure, but it sounds like you are not dealing with RGB images, so don't put that into 8th dim.

You can simply use the 4th dim to store the x/y/z-deformation component, as FSL warping files do. Although NIfTI specification reserves 4th-dim for time, but no one cares or follows it.

I suppose the metadata will be same as or similar to whatever NIfTI you derive the deformation vectors from.

[adityaapte]

When I save x/y/z deformation in the 4th dimension, the image written to nii is 4-D. It has 16-element direction cosines (instead of 9 element). Also, NumberOfComponentsPerPixel=1 instead of 3. Perhaps I am missing the metadata required to tell the nii writer to treat the 4th dimension of the image as 3 components per pixel instead of a separate dimension.