I know this may not be the ideal place to post this since it is more of an EDDY issue, but I was hoping that someone has encountered something similar and found a solution.
I’m having problems with preprocessing diffusion data. dwifslpreproc runs without errors, but the results do not look correct.
First, the images have stripes (possibly slice-wise translation due to multiband). After preprocessing, the eddy QC output graphs look strange to me. Below in the eddy estimated translation graph, the translation in “y” direction looks weird and also translation in the range of 8 mm (across multiple subjects, not just this one) seem too large. Also the stripes remain in the data after preprocessing. Below, I have also attached other graphs from eddy QC.
Has anyone encountered a similar issue? Could eddy be processing the data incorrectly, or is this possibly a hardware problem?
It might be useful to show screenshots of images before and after dwifslpreproc to better demonstrate the effects you are describing.
Particularly for “stripes”, it’s important to distinguish whether such effects are present in the input raw data vs. arising specifically from the processing of this command: the former could be inadequacies of the acquisition, whereas the latter could be caused by something like an incorrect allocation of slices to multi-band slice stacks. Showing the output of mrinfo on the input DWI, and the resulting slspec file that is generated and passed to eddy, might be useful.
For assessing the quality of pre-processing, low-resolution GIFs that loop across volumes—one for pre-dwifslpreproc, one post—can be useful to contrast whether there’s something fundamentally going wrong in the estimatino of subject motion.
