Hi everybody!
I have a Situation where I want to apply a nonlinear deformation (a warp) from a registration to a Diffusion weighted Image (DWI). In order not to depend on gradient directions (and for further processing) I converted the DWI into its spherical harmonics (lmax=8) representation using amp2sh. I know that there is an option “-reorient_fods” for mrtransform, that ensures the correct handling of the FODs when transformed.
My question is if this is also feasible for the spherical signal components prior to CSD. Results look ok so far. Given that CSD makes some assumptions (non negativity …) what could be a practical/numerical limitation here?
Thank you very much for any advice/discussion about this.
Fabian