Can I ask what that means and how to fix it?

[xuerange]

Sag_DKI_25dir_2B: bvec sign for non-axial acquisition with ROW phase direction not tested.
Please check the result and report problem to author.

[xiangruili]

This should be a diffusion scan from GE, right? Most people use axial slices, so that is well-tested. For non-axial slices, we have very limited testing dataset, so the sign of the bvec might be wrong for GE system. You can test the result, let us know if it is right, or you can share the dataset with us, the converter developers, so we can test it.

[neurolabusc]

@xiangruili this seems timely. You may want to see the GE validation dataset and dcm2niix bvecs with this PR.

[mr-jaemin]

I tested non-axial acquisitions (with dcm2niix) and didn't experience any issue so far. Here is some validation dataset:

exam5382_multishell_orientation_invivo (1.6 GB)
See README_multishell_DTI_Orientations_Ex5382.pptx
Axial/Sag/Cor various orientation including
Strange oblique suggested by Chris

As @neurolabusc noted above, I found that feet first DTI works the same way as head first.

[xiangruili]

The warning "bvec sign for non-axial acquisition with ROW phase direction not tested" is not because of non-axial acquisition, but also InPlanePhaseEncodingDirection="ROW", which is rare.

The validation dataset all have "COL" phase encoding direction. For manufacturers other than GE, this does not matter. But for GE, this could affect sign for bvec.

@xuerange If you can share your "ROW" dataset, preferably with some oblique angles, we could verify how GE dicom encodes bvec sign.

[xuerange]

@xiangruili Thank you very much for your reply, I tried to send you the data yesterday but it didn't work, I don't know how to upload the data, I had compressed the data and tried to drag it directly to the dialog box but it prompted me that the file was too big to upload!

[xuerange]

@xiangruili May I ask Prof. Lee to give me an e-mail address?

[xiangruili]

@xuerange My email is [email protected] . But email attachment likely won't work. You should share it with a temporary OneDrive folder or something alike.

[xiangruili]

@neurolabusc
I just compared the bvec from dicm2nii.m and dcm2niiX (v1.0.20220720). The FF version gives different bvec, but I think I was using older version of dcm2niiX. In dicm2nii.m, I don't do anything differently for HF vs FF, and that is what you verified, right?

[mr-jaemin]

@xiangruili Yes, the bvec conversion for FF would be same as HF based on the validation dataset.