QSIPrep DWI and T1w registration Dims don't match


I am running QSIprep on a test subject and have run the initial preprocessing pipeline. I have two questions.

  1. When comparing the dims for ses<>/run-01_space-T1w_desc-preproc_dwi.nii.gz and _desc-preproc_T1w.nii.gz the dims are as follows:

ses<>/run-01_space-T1w_desc-preproc_dwi.nii.gz: dim1=195, dim2=232, dim3=197, dim4=127
_desc-preproc_T1w.nii.gz: dim1=193, dim2=229, dim3=193

Why are the dimensions not equal after registration?

  1. The T1 is flipped after running qsiprep. I am currently running version 0.15. I see that 0.16 is available and am planning to run qsiprep after building the singularity image using 0.16 version. Will this resolve the flipping issue?


Hello Tejaswi,

QSIPrep does not resample the voxel size of your DWI image to the T1.

In what way is it flipped?


Thank you, Steven. I guess QSIPrep is similar to SPM.

The t1w is flipped in orientation. I ran fslreorient2std and ran qsiprep again but it flips the orientation after the T1w is reoriented.

To resolve this issue, I ran using qsiprep 0.16.0R3C. but this resulted in the T1s being completely off (only half of the brain shows up on fsleyes).

In what orientation is the image flipped?

What do the preprocessed T1 and DWI images look like overlayed onto each other (before you did any reorientation)?


Hi Steven,

The T1ws are flipped from anterior to posterior post qsiprep processing. The T1w and DWI images correspond well with each other when overlayed but their dimensions do not match because they are not re-sliced.

That is probably just a coordinate system convention. As long as the direction labels are correct (that is, the anterior label matches the anterior part of the brain and so on), nothing to worry about.

That’s all that matters. You probably would not be gaining anything by interpolating your DWI to whatever high resolution your T1 was collected at; at some point there is only so much signal.


1 Like