Does convert_sh_to_legacy work as intended ?

[DelinteNicolas]

Description

I was converting ODFs files, saved using spherical harmonics (SH), from the MRtrix3 format (Tournier) to DIPY (Descoteaux) and had issue with this function. It currently flips the SH coefficients when m is odd and <0 : out_sh_coeffs = sh_coeffs * np.where(m < 0, (-1.)**m, 1.).
However, I noticed that the coeffifient that I needed to flip to get a working files were the 'even' m numbers. As mentionned in the DIPY theory (https://dipy.org/documentation/1.4.0./theory/sh_basis/) : "the effect of this change is a sign flip for the SH functions of even degree m<0".

If I understand correctly, shouldn't the code line be replaced with out_sh_coeffs = sh_coeffs * np.where(m < 0, -1*(-1.)**m, 1.) ?

With:
DIPY==1.5.0
MRtrix==3.0.2

[skoudoro]

Yes, it worked as intended.

In your case, you might want to use convert_sh_to_full_basis instead.

@CHrlS98 can tell you more when he has time.