Pipeline Tips

Dear experts,

I am implementing a pipeline to preprocess and analyse our lab images. We recently updated our protocol to a MultShell DWI with 10b0, 24b1000, 32b2000 and 56b3000 with a separated additional B0 image (NSA=10) with inverse phase-encoded direction.

In general, my pipeline is based on:

  • DWI-denoise
  • Gibbs Ringing correction (AP and PA images)
  • dwifslpreproc (-rpe_pair -se_epi “b0_pair.mif” -pe_dir ap -eddy_options " --slm=linear --repol" -align_seepi)
  • dwibiascorrect ants
  • dwi2response dhollander (-shells 0,1000,2000,3000)
  • dwi2fod msmt_csd
  • mtnormalise
  • tckgen wmfod_Norm.mif…

Is this a good approach, or should I consider different algorithms?

Thank you very much!

If you have enough storage and computation capabilities I suggest upscaling the images inplane before dwifslprerproc. This will give you much sharper data. It noticeable to the naked eye.
Don’t upsample the slices, as that will break eddy correction.

1 Like

Hi Mark, thanks!

Are you suggesting a 2D interpolation over the acquiring plane, making the voxel anisotropic? We acquire and reconstruct 2x2x2 mm3, you suggest something like 1x1x2 mm3 or less?