 |
 |
I have some multishell data which I've been processing with eddy/topup. Due to limitations, the multishell data was acquired as single shell scans (all in the same scanner session; with the same imaging parameters), concatenated, and then processed with eddy/topup. This worked fine until recently. One of the scans in the set I'm processing actually has a different TE than the others so the overall intensity is higher. I'm thinking this will present problems with eddy as it does it's "fitting" on a per-shell basis if I'm not mistaken, and I now have two b1000 scans and one has an overall higher intensity than the other. I can't normalize the scans by dividing by the b0 first, because that would require motion correction first, but I want to do motion correction with eddy, so this provides a somewhat circular dependency. Is there a best practice for this? I know I can eddy/topup scans individually and then register the b0s together with a 6DOF flirt, but I'd like to avoid this. I'd like to concatenate all the scans and put them through eddy/topup. Do you think I can get away by dividing each scan by the median of the b0 (thus setting each scan's median b0 value to 1)? Then afterwards split them up and then divide by each scan's b0 to do the "real" normalization? This isn't quite "right" but I have noticed that I get much better results by concatenating scans and eddy/topuping them than eddy/topup individual scans and registering them together.