In versions 0.13 and later, QSIPrep finds the “most representative” b=0 images per warped group.
A nearly identical approach is used in the developmental HCP pipelines, where a pairwise spatial correlation score is calculated between all b=0 images of the same warped group and the images with the highest average correlation to the other images are used as input to TOPUP.
To see which images were selected, examine the selected_for_topup column in the confounds tsv file.
I don't see selected_for_topup in my confounds using 0.14.3. I tested on ABCD data but others have seen the same issue with HCP data.
I couldn't find it in the selected_for_topup column in the confounds tsv file, but I think that the TOPUP Inputs section in the ..data/derivatives/qsiprep/sub-XXXX.html can be helpful.
From the docs:
I don't see
selected_for_topup
in my confounds using 0.14.3. I tested on ABCD data but others have seen the same issue with HCP data.Here's the command I ran