Wondering about how to interpret some of the QSIPrep html outputs or find resources about acceptable ranges for quality control purposes, including the following:
SHORELine Convergence violin plots: Is there a reference range or generally acceptable limit for convergence measures (beyond being close to 0)?
Framewise Displacement: Similarly, is there a generally acceptable limit/ range for diffusion weighted imaging, given inherent spikes at b0 timepoints? Would the range be similar to acceptable limits for fMRI in between the spikes?
Many times this depends on your study design. Dense acquisitions tend to be more robust to motion spikes and moderate average motion because there is a lot of redundancy in the DWI signal between nearby directions, especially when fitting low-dimensional models like the diffusion tensor.
So you might be better off, beyond excluding visual inspection of the images from the HTML and/or by loading the dwiqc.jsons into dmriprep-viewer, using a QC metric like neighboring correlation as a covariate in your statistical models.
I do not know enough about SHORELine to answer that first question.
Hi @ellem! The SHORELine convergence plot was added in a very long time ago when we were first developing SHORELine and we can probably take it out. It really is just to make sure that the values are near zero.
As for the framewise displacement, I don’t think dMRI should be treated like BOLD, where there is a standard cutoff for framewise displacement. Head motion, eddy current correction and dropout imputation tend to not to have as big of effects downstream.
I agree with @Steven that measures like neighbor corr are more useful for inclusion/exclusion decisions.