SynB0 issue

1

Hi all,

I’m using SynB0 and dwifslpreproc #2266. After running SynB0, the topup files were fed into dwifslpreproc #2266 as follows, which ran without error, but it seems distortion still visible in the output files.

dwifslpreproc Diffusion_denoise_unr.nii.gz Diffusion_denoise_unr_eddy.nii.gz -rpe_none -pe_dir AP -fslgrad bvec.bvecs bval.bvals -eddy_options=" --topup=topup"

The outputs of synB0 (distortion was corrected)
Screenshot (20)

The output of dwifslpreproc (seems still distorted, but differs a bit from the eddy output if I don’t feed any topup file)
Screenshot (19)

Did I miss some steps? Any help would be highly apreciated!

2

Hi @XL258W,

There’s too many variables at play here to diagnose where a problem may be, if indeed there is one; so it would require more experimentation.

E.g.:

  • Is there an error (relative to reality) in the inhomogeneity field estimate, which is not visible when considering only the b=0 data but when examining the DWI data it’s clear there’s a residual error?

  • Is there an error in estimation of motion in that DWI volume, such that the inhomogeneity field correction is applied incorrectly?

  • Is the output of SynB0 being interpreted incorrectly by dwifslpreproc? What happens if you pass the output of that tool directly to eddy and bypass MRtrix3 entirely?

Rob

3

Thank you very much for the help, Rob. I will check those points!

4

@rsmith
Thank you for continuous improvement of dwifslpreproc.
Since the 3.1 is not yet released, Is there any good recommendation to adopt the result of synb0disco result to dwifslpreproc? It seems one of the best option for recover distortion of single-direction-only DWI image.

5

The dev branch of MRtrix3 has a -topup_files option that allows you to specify the synb0 outputs. It’s a bit tricky as you will need to compile the dev branch from source vs. the more convenient installation strategies.