I need an opinion here, and maybe @jdtournier or @rsmith to chime in. I’m working on mouse data and the dubious conversion to NIFTI and all of the challenges associated with it.
In the current workflow, I know that the b-vecs are not oriented correctly and that dwigradcheck can give me correct ones. However, as these are ex-vivo, there is occasionally fairly substantive motion (e.g., brain sinking in field of view). Here’s the think I’m trying to reconcile in my head:
- If I run
dwigradcheckearly in my pipeline (e.g., after denoising and unringing), then I may get less correct output because of gross misalignment of the brain over the course of the accquisition. - If I run
dwigradcheckafter motion correction, then my bvecs have already been reoriented bydwifslpreprocbut done so technically incorrectly (e.g., if major correction occurs in the y-direction, then this will result in changes to the y-direction of the vector as it is read in but if I really need a 102 swap then I’ve actually adjusted the x-direction of the vector)
Am I just overthinking this?