Thanks for your help, everyone,
Looks like there's no out-of-the-box solution, so I'd like to run a proposed fix by you guys (inspired by Saad's suggestion). We have a pretty wiggly cohort (ADHD kids), so we wanted to keep as much data as we can. Our processing stream is set up with RESTORE (Chang et al - Mag. Res. Med. 2005), for tensor estimation. RESTORE has been working really well for us.
So how about this: I could patch artifacts in the original DW images by replacing them with values generated from the RESTORE-estimated tensor, then feed those patched DWIs into bedpost. I'm reasonably confident that RESTORE doesn't just toss out deviations from the tensor model due to crossing fibers.
What do you think, does that sound reasonable?
-Dan
|