This fixes the problem: dwis look good, fODF can be computed from them and the speed is fine.
I guess the problem is the same (1/(4 * np.pi)) in compute_fod_sh.
The 1/(4 * np.pi) makes the diffusion value very low compare to the b0 value and i guess this makes the fODF computation (dipy) fails.
Thanks Gab. This really calls for the development of testing of the code of some point. I will try to start putting the basis for a propser test-driven development.
This fixes the problem: dwis look good, fODF can be computed from them and the speed is fine. I guess the problem is the same (1/(4 * np.pi)) in compute_fod_sh.
The 1/(4 * np.pi) makes the diffusion value very low compare to the b0 value and i guess this makes the fODF computation (dipy) fails.
What should be done?