Dear Jonathan,
> I have a dataset of DWI images acquired only in one direction and few weeks ago I asked about EPI distortion correction in this forum using a T1-weighted image with ANTs.
>
> I finally ended using BrainSuite with, which I think, pretty good results (I attached a jpeg for comparision)
>
> However, since I haven't two datasets in opposite directions, I only can apply eddy_correct instead of "eddy" which I believe it rotates bvecs automatically. After that I need to rotate my bvecs with the Mark Jenkinson's script (fdt_rotate_bvecs).
I think you have misunderstood that. It is fine to run eddy without opposing directions (whether these are PE-directions or diffusion gradient directions). And, as you say, eddy will automatically rotate the bvecs for you.
If you run eddy without supplying it with a susceptibility field (ideally in the form of topup output, but you can also input a fieldmap (in Hz)) it will still correct data for eddy currents, subject movement and signal loss, but not for susceptibility.
I am not familiar with the details of BrainSuite, but if you can get it to output a displacement field in the space of the EPI image you can then rescale that to Hz and input it to eddy using the --field option. If that works, I think that would be the easiest solution. I hesitate to give advice on multiple rotations of the bvecs with sign flips thrown in. What we say about the rotation in eddy is that _if_ the bvecs that are put into eddy are in the FSL convention then the rotation is correct. To check that the bvecs are in the FSL convention the easiest is to run your data through dtifit, look at them in fslview or fsleyes and make sure that the first eigenvectors line up across voxels.
Good luck Jesper
>
> I'm confused about bvecs rotation because BrainSuite gives already as an output an bvec rotated file after align the DWI to the T1 sequence.
>
> Therefore I have the (1) the original bvec file (CA043_DTI.bvecs), (2) the rotated bvec after eddy_correct (rotated.bvecs) and (3) the rotated bvecs from BrainSuite (CA043_3D.dwi.RAS.bvec). I attached an example of the outputs, since BrainSuite changes the sign of all axis and I'm kind of insecure about this.
>
> My pipeline for preprocessing was (1) eddy_correct (2) rotate_bvecs (3) skull-strip and bias field correction (BET, FAST) of the T1 (4) BrainSuite (bdp.sh -T1.bfc.nii.gz --nii dwi_eddy.nii.gz -b rotated_bvec -g bvals) and (4) BET b0 and (5) DTIFIT using the bvecs from BrainSuite.
>
> I'm having different results in TBSS since I've done this type of preprocessing.
>
> I appreciate any kind of suggestions/commentaries. I'm really interested in running a TBSS analysis with the EPI distortions solved and the bvecs reoriented.
>
> Thank you very much,
> Jonatan
>
>
>
>
>
>
>
>
>
>
> <ejemplo_EPI_brainsuite2.jpeg><CA043_DTI.bvecs><rotated.bvecs><CA043_3D.dwi.RAS.bvec>
|