A note about the output units of 3dDWItoDT:

+ 3dDWItoDT will now use in the input units of the b-values and give output diffusion coefficients that have matching units. So, if your input b-values are ~1000 s/mm^2, then your output MD, diffusion tensor values, L1, L2, etc. will have physical values of 'mm^2/s'. Great.

+ In the human adult WM, typical values for MD/diffusion are around 0.7 x 10^{-3} mm^2/s, so that the output diffusion-united values may likely have lots of leading zeros: 0.0007 mm^2/s, etc. This might be an annoying feature for people who would rather not have lots of zeros, so there is a new option in 3dDWItoDT to multiply the diffusion-united outputs by 1000, effectively meaning that the units are "10^{-3} mm^2/s", instead of just "mm^2/s". This is the description:

-scale_out_1000 = increase output parameters that have physical units
               (DT, MD, L1, L2 and L3) by multiplying them by 1000.  This
               might be convenient, as the input bmatrix/gradient values 
               can have their physical magnitudes of ~1000 s/mm^2, for
               which typical adult WM has diffusion values of MD~0.0007
               (in physical units of mm^2/s), and people might not like so
               many decimal points output; using this option rescales the
               input b-values and would lead to having a typical MD~0.7
               (now in units of x10^{-3} mm^2/s).  If you are not using
               bmatrix/gradient values that have their physical scalings,
               then using this switch probably wouldn't make much sense.
               FA, V1, V2 and V3 are unchanged.

+ This is really a matter of convenience, and because with very small numbers, some of the statistical modelling or further calculations may run into numerical issues with small differences of small numbers going below floating point precision, hence returning 0s when they probably don't need to.

+ NB: the scalar FA values and eigenvectors V1, V2 and V3 are all unaffected by this change, as they are all unitless things-- FA is a scaled to be [0,1], as always, and the eigenvectors are just unit normal direction-pointers.


In summary, this new option is just there as a matter of convenience, not to change the meaning of any output, but to save one from artificial numerical problems due to having tiny numbers in calculations.

--pt