Hello
The DTI estimation in the diffusion workspace is not yet taking into account b-values. In fact, it acts as if the b-value was one. In terms of tensors, it doesn't change anything as long as you have an acquisition with only one b-value. In that case, you may just divide the ADC values or the tensors themselves by that b-value to get the real ones. However, in its current status, medInria won't estimate correctly tensors from acquisitions with multiple b-values. This will come in the future.
As for the gradients, it is better to have them normalized, although I remember they are probably normalized inside the algorithm but you are never too careful. Finally, the check-box you notices is here to handle different gradient orientations from the scanner. In fact, depending on the scanner, the gradients may be expressed in the image referential or in the real world referential. We consider by default that the gradients are in image referential, but if you see any wrong behavior with that setting, you can check the other option to solve the problem.
Hope this helps
Olivier