Hi-

Different programs use different orderings of columns of the bmatrix (row- or diagonal first; also, there's a difference of a factor of 2 in the off-diagonals). Assuming that you are using the correct orderings for a given program, then the gradient and bmatrix are mathematically equivalent.

The difference could involve the 'flipping of the gradients' difficulty... In different softwares, one is required to flip the sign of a gradient sometimes in order to have the correct vector directions (the scalar quantities FA, MD and eigenvalues are not affected by this)-- for example, flipping the y-component of the gradient flips the y-component in each of the eigenvectors. In the *newest* TORTOISE (v2.0.1), which is what was used at the Bootcamp, Okan's data didn't need to be flipped. However, if you have data from a different scanner, then it might need to be flipped. As far as I know, this is just the way things are. You can tell if you gradients are fine or if the particular flip is fine if, for example, a whole brain tractography test gives 'correct' looking results along the well-known corpus callosum.

It sounds in your case that this is not the case with the TORTOISE results, and so you should flip one of the signs of one of the components in your TORTOISE-output eigenvectors, for example the y-component, which is hte [1]-th brick of each vector.

To answer your questions:
1) You should be fine using either the direct estimates of TORTOISE data or doing the fit with 3dDWItoDT. For the processed data, the results should be *very* similar, particularly in WM, differing mostly in ventricle/CSF/boundaries between tissues a bit.

2) Grad/Bmatr/Gmatr are equivalent. You can use 1dDW_Grad_o_Mat to do flipping in the gradients or matrices equivalently as well. If you have reconstructed vectors already, then you can flip a given element/component using 3dcalc.

Please let me know if your issue persists, but flipping should do the trick, I'd imagine, for hte TORTOISE output. Once you have found the correct flip, then that shoudl apply to the rest of your data from that same scanner, hopefully.

--pt