Does the MRICron-generated NIFTI look as expected in mrview? (see Overlapped ROIs image and diffusion images)
Assuming a negative answer, I’d guess the issue is that the sform and qform convention in MRICron is different from what we use (see: NIfTI transform ambiguity). You could instruct MRtrix to ignore the qform via the config file option NIfTIUseSform or command-specific via command-line option -config NiftiUsesSform 1 to see if that resolves the issue.