Hello AFNI Wizards!
Apologies if this question has been asked (I did try a quick search):
I am attempting to perform some basic probabilistic tractography with 3dTrackID. I know for 3dDWItoDT, the gradient vectors file should omit the G0 line. For 3dDWUncert, should the gradient vectors file (-grad option) also omit the G0 line, or should that be included for this program?
Thank you in advance!
~Nate
Hi, Nate-
My preference for both 3dDWItoDT and 3dDWUncert is to use “-bmatrix_FULL [in_file]”, where [in_file] is a set of AFNI-style bmatrix values. In this case, if you have N total b0+DWI volumes, then you also use all N associated bmatrix values (equivalent to grads, just the outer product). I really don’t like having to worry about chopping off the [0]th grad.
Also, I strongly prefer having either the grads or bmatrix being scaled by the b-value weighting, and then use “-scale_out_1000” when you run 3dDWItoDT: this will leave the MD, RD, L1, L2 and L3 values all in physical units: “10^{-3} mm^2 / s”. That way the numerical values should be of order 1 or so (and having possibly several different bvalues is still accounted for correctly). It might help with statistical modeling later, so the numerical part of the MD values aren’t very small (and taking differences of small numbers might lead to rounding errors). NB: 3dDWUncert does not need this option, beccause it is only calculating uncertainty of FA and V1, both of which are unitless.
To convert grads to bmatrices or anything, please use 1dDW_Grad_o_Mat++. An example for you to simply convert a full column file of weighted grads to an AFNI-style bmatrix would be:
1dDW_Grad_o_Mat++ \
-in_col_vec GRAD.dat \
-out_col_matA BMAT_matA.txt \
If your grad file is unit magnitude and the bvalues are in a separate file, then you could do:
1dDW_Grad_o_Mat++ \
-in_col_vec GRAD.dat \
-in_bvals BVALS_file.dat \
-out_col_matA BMAT_matA.txt \
If you need to flip directions, you can do that with the above function.
If you need to check if you need to flip, please see/use @GradFlipTest.
–pt
Dr. Taylor,
Thank you very much! This is very helpful and exactly what I needed!
Cheers,
~Nate