Thanks for your answer,

I am sorry I didn't mention. I am already working with scalar images in my case just FA and AD (axial diff.).
However the problem is that the scalar images present different levels of FA (after watching the histogramms), depending on the scanner, and I don't think that a straight forward comparison is convenient.

I have been recommend to assign some scaling factor over the healthy sample to the scanners, in order to put them all in the same normalized space, so I did it.

However I was interested in knowing if any software deals with interscan variability for this kind of analyses.

Thanks
Jose