Thank you for your response!
I guess my question was also about tensor interpolation. In other words, how
does FSL reconstruct a continuous tensor space from the discrete sampling
provided by the number of the gradient directions. Also, when the tensors
are fit according to the eq from Basser 1994b, how does FSL ensure that when
the model is fit to data, the resulting symmetric matrix is a tensor, that
is the eigenvalues are positive.