Firstly, I would recommend having a look at the FATCAT demo set, or at
least the overview README.txt file, for various possibilities and
pipelines of analysis. It is easily downloadable from here:
[
afni.nimh.nih.gov]
and it contains lots of scripts to base your analysis on.
Since you have ROIs in each subject's diffusion space, you are ready
to go. Any set of regions that you are interested in finding
connections among should be put into a single AFNI brik, each with a
different integer label. If your ROIs are currently separate masks
(roi_01+orig, roi_02+orig, etc.), then you could combine them as
suggested in this thread:
[
afni.nimh.nih.gov]
You are now ready for tracking. You do not need FSL-bedpostX results.
To run probabilistic tracking, which sounds like you want for your
data, you will first want to run 3dDWUncert to estimate the
uncertainty of relevant DTI parameters in your data. (As a note, I
would probably use 3dDWItoDT to reconstruct your tensors instead of
dtifit-- the latter takes the absolute values of eigenvalues that are
negative instead of refitting the tensor, and gives voxels with FA>1,
neither of which are very good things; it is a quick thing to rerun.)
Once you have your tensors and 3dDWUncert results, you can perform
probabilistic tracking using '3dTrackID -mode PROB ...' This is how
you only include voxels in your tracked WM regions that have a lot of
tracks going through. You specify the fraction you want as threshold
(for example, your 0.05 from above) when you run the command, and the
thresholding occurs internally.
The output of tracking will be: maps of WM regions connecting your
targets, as well as matrices of FA, MD, L1, etc. values for each
region (the mean and standard deviations). You don't need to do
3dcalc separately. Also, when using 3dTrackID, I usually use the
'-dump_rois AFNI' option so I have distinct maps of every region
output-- I find that useful.
In terms of visualization, SUMA is a good bet as well. You can view
probabilistic tracts as volumes (-vol NAME).
In addition to the above, I would recommend using the
mini-probabilistic tracking for viewing 'tract-like' results that also
take into account voxel uncertainty. This can be done with '3dTrackID
-mode MINIP'... SUMA will also be your friend for this.
Executive summary: many of these steps are laid out in the
FATCAT_DEMO, and I can't recommend enough taking a look through the
comments and running the pre-built scripts there.
--pt