Dwipreproc on multiband (HPC-like) DSI failed


When I run dwipreproc following the latest instruction on multiband (similar to HPC) DSI, I got an error (I used reconstructed nitfi images, instead of raw pre-reconstructed DICOM).

dwipreproc: [ERROR] Command failed: eddy --imain=dwi.nii --mask=mask.nii --acqp=eddy_config.txt --index=eddy_indices.txt --bvecs=bvecs --bvals=bvals --topup=field --out=dwi_post_eddy (dwipreproc:518)

the error.txt says:

eddy --imain=dwi.nii --mask=mask.nii --acqp=eddy_config.txt --index=eddy_indices.txt --bvecs=bvecs --bvals=bvals --topup=field --out=dwi_post_eddy
Image Exception : #3 :: Attempted to add images/ROIs of different sizes

Any suggestions? How do you preprocess multi-band DSI using mrtrix?

Hi @Jiook_Cha,

Two separate parts here:

  1. The “Attempted to add images/ROIs of different sizes” error was reported here also. That is not an issue of either multi-band or DSI, but of one or more of your spatial axes having a number of voxels that is an odd number. My initial work-around for this issue in dwipreproc does not work with newer versions of eddy, so you will need to manually crop / pad your images before running dwipreproc.

  2. Even after correcting the above, you may yet still encounter an issue such as this one. Even if dwipreproc succeeds (which will depend on how stringent eddy chooses to be`, there’s not much you can do in MRtrix3 with DSI data; the software is largely built around the spherical deconvolution model, which requires q-vectors arranged in spheres rather than on a grid as in DSI.