Incorrect/incomplete whole-brain tractography result

[Maria Economou]

Hi all,

I'm using a Philips multi-shell dataset with 119 volumes, pre-processed in FSL (eddy, motion and susceptibility distortion corrections + dt fitting). I then used the function dtiMakedt6FromFsl to create the dt6.mat file.

I ran the AFQ pipeline on several (>100) subjects and after checking the output I noticed that for a small subset of them (but not all) most of the fiber groups are empty and the whole-brain tractography results look quite strange/incomplete (see attachment).

I already double-checked the acpc-alignment and image quality for both the diffusion and T1 datasets and everything looks fine. By going through some older posts, I suspect it could have something to do with the gradient direction table, however I cannot tell what is wrong by looking at the first eigenvector image (see attachments for the 3 planes). A solution offered to that problem was to flip the gradients over the x-axis -  how can I confirm that this is indeed what's wrong? And shouldn't this be the case then for all subjects processed with exactly the same pipeline? 

I'm baffled by the fact that for some subjects it all seems to be working fine but for others not. Any ideas on which step could be causing the issue (and possible ways to solve it) are greatly appreciated! Please let me know if I should provide additional info.

Many thanks,

Maria