I tried 1) dicom to mif conversion using mrconvert and 2) dicom to nii and then nii (+bvec/bval) to mif on the same subject.
I ran a pipeline that includes denoising, gibbs removal and then concatenated them (2 acquisition directions) and then performed dwifslpreproc for both the above 1) and 2).
After that I got the output of both and ran DTIFIT, 2) produced a nice FA image, but 1) produced a very messy FA image. All methods after the first file conversion step are the same, so I was surprised to see that. I see this in the FA, MD, V1, etc, otherwise the images look very similar.
Any insight is much appreciated?
PS: I checked the bvacs files output after dwifslpreproc and they are different (in values and sign).