I am relatively new to neuroimaging research and I am setting up a diffusion preprocessing pipeline (planning to use QSIprep). I have a technical dilemma regarding Susceptibility Distortion Correction (SDC) and what would be considered the most scientifically rigorous approach for a future publication.
We acquire a high-resolution multishell DWI sequence (ABCD protocol). Unfortunately, we do not acquire a dedicated reverse phase-encoded b0 associated with this exact sequence.
pixdim1
0.9766 mm pixdim2
0.9766 mm pixdim3
2.5 mm pixdim4
5.8 seconds
We do acquire a DWI MUSE sequence in the same session, which includes a reverse phase-encoded b0. However, this MUSE sequence has a significantly lower spatial resolution compared to the ABCD multishell.
pixdim1
1.016 mm pixdim2
1.016 mm pixdim3
4 mm
My dilemma: I have run some preliminary tests using the lower-resolution MUSE b0 to correct the ABCD multishell (via TOPUP / DRBUDDI) and, upon visual inspection, the results actually look quite reasonable. However, I am unsure what is the standard practice or most acceptable method for a manuscript.
My questions are:
Is it scientifically sound to use a lower-resolution reverse-phase b0 (from a different DWI sequence) to correct a higher-resolution multishell acquisition using TOPUP or DRBUDDI?
Or, given the resolution mismatch, would it be more recommended to skip the MUSE b0 entirely and use a fieldmap-less algorithm like SyN (using the T1w anatomical image)?
Any advice on the trade-offs between using a mismatched-resolution b0 vs. fieldmap-less SDC would be greatly appreciated.
It is hard to me to prescribe exactly without having more hands-on exploration with your dataset, but for what it’s worth, the official ABCD processing pipeline uses SynB0-DisCo, even though fieldmaps are present. We are still in the process of incorporating a fieldmapless SDC method into QSIPrep to replace the now outdated --use-syn-sdc method. Personally, I would opt to use SynB0, or at least try it and see if it looks reasonable for you.
But also,
That doesn’t sound like an ABCD DWI sequence (which is 1.7mm isotropic).
Thank you so much Steven for your quick and helpful reply. I will explore SynB0-DisCo for our dataset.
Regarding the sequence resolution, I inherited this dataset and I am not sure if our MRI technicians adjusted the original ABCD parameters to fit our scanner’s time constraints/coils, or if it is actually a completely different sequence that was just named “ABCD”.
I have some of the parameters from the .json file of our multishell DWI sequence. Could you take a quick look and let me know if this looks like an adapted ABCD sequence or something else entirely?
Thank you so much again for your time and answer .
Your protocol seems to use different TE, TR, and voxel resolution. The TE and TR might be normal for adapting to a different scanner, but the voxel resolution change is surprising. Seems to supersample the axial plane while sacrificing a lot of z-axis resolution. Usually DWI is acquired at isotropic or near isotropic, and these high in-plane resolutions are usually more common in clinical settings. The gradient table might be similar to ABCD, given the 102 directions, but confriming that would come from the bval/bvecs, not this. In either case, the specifics are not too important, but I would be cautious describing this as an ABCD acquisition in the future!
.