Unsure about differences in preprocessing of volumes used for topup

[psadil]

I'm working with a dataset that has dedicated fieldmaps collected with reversed phase encoding directions. It's my impression that the topup node, when in the pipeline, builds the input from combinations of the dedicated fieldmaps and b=0 volumes from the DWI (currently without considering the --prefer-dedicated-fieldmaps option, as per #321). However, when QSIprep is configured to cleanup the DWI scan through denoising, ungrining, and bias correction, it seems like that cleanup only happens for the volumes sourced from the DWI scan itself, and not the fieldmaps. So, the inputs to topup comprise both raw and preprocessed volumes.

Would it be feasible to avoid this difference, so that all volumes input to topup receive the same treatment? For example, two ways to do this would be to restore the functionality of the --prefer-dedicated-fieldmaps option (so that all inputs to topup are raw), or to also preprocess the fieldmaps prior to topup.

Thanks!

QSIprep run command

singularity run -e --nv docker://pennbbl/qsiprep:0.14.2 \
  "${BIDS}" . participant \
  --output-resolution 1.7 --bids-filter-file bids.json \
  --denoise-method patch2self --unringing-method mrdegibbs \
  --hmc_model eddy --eddy_config eddy_params.json \
  --fs-license-file license.txt --participant_label "${subject_id}" -vvv​

[mattcieslak]

Sorry for the super late reply - as of 0.16.1 the images used for topup are forced to have the same level of processing. This ended up being important for ABCD data

[psadil]

That's great! Thanks for the response (and patch). Just to be clear, does this mean that these updates will not be available in the version that currently has a release candidate, 0.16.0RC3, but that instead I should expect to see them in the subsequent version?