I am working with some complex volumes that need registering. My go-to method for this is running
mcflirt on the magnitude volumes, splitting the complex volumes into real and imaginary parts, using
applyxfm4D on these separately, then re-combining (this avoids problems with sampling phase across the 0/2pi boundary).
What is the correct way to pass the output matrices of
ApplyXfm4D takes the
trans_dir argument, because the underlying
fsl tool also takes a directory as an argument that then contains all the transformation matrices. However the
nipype wrapper of
mcflirt does not expose this directory as an output, instead it provides a list of all the individual transformation matrix files.
For now I’ve hacked the
MCFLIRT wrapper to also provide the output directory, but I feel like I’m missing something and ended up re-inventing the wheel.