On Siemens scanners, ep2d_bold sequence has a motion correction option (the retrospective version, so not what they call “PACE”). I wanted to hear the opinion of the fellow neuroimagers here and whether people recommend using these series or not.
My worry is that if we use the motion corrected series, the corrected series would lack necessary parameters which can be used as covariates for GLM. So I think instead of doing online correction, I shall do offline correction with full control under my hand?
We had an issue with MoCo series where 1 volume has completely disappeared from the scan (the volume is there, but there is no brain tissue) at the subsequent scans, the brain is back. But after a few volumes, 4-5 slices from the top side have completely disappeared and after that final volume, the motion corrected series were not generated at all (uncorrected version is intact) so I am curious if anyone else experienced such a bizarre thing.
Has anyone extensively compared these two versions when participants undergo excessive motion?
AFNI, FSL and SPM all have their own methods for generating motion-stabilized images. I would use these instead of the vendor supplied version for three reasons:
As you mention, the analysis stage is aware of the motion and you model it as a nuisance regressor.
One can detect extreme motion to decide if the data should be censored completely or via motion scrubbing.
Preserving the complexity of the source data can be useful for future re-use and reproducibility. Consider someday a future tool that integrates motion and slice timing into a single stage (as these interact with each other).
I agree with all what @neurolabusc wrote. In your case, I would also recommend to go for offline correction for the same reasons.
Yes, we observed what you noticed with the moco series: indeed, once a specific threshold in the motion parameters calculated by the moco series is hit (2mm in translation or 8 degrees in rotation, I don’t remember exactly) the MOCO series does not reconstruct anymore, even if the uncorrected series does. Also the Siemens console sends you a warning (a yellow crossed symbol of a RF wave in the bottom of the acquisition screen, telling you that those specific motion thresholds were hit).
Otherwise, since you mentioned the PACE-fMRI sequence, I indeed noticed that some people successfully used this sequence for fMRI, I put the references here:
(1) Lanka, P.; Deshpande, G. Resting State FMRI Data from Subjects Scanned with the EPI-PACE (Echoplanar Imaging – Prospective Acquisition CorrEction) Sequence. Data in Brief2018, 20, 2072–2075. https://doi.org/10.1016/j.dib.2018.01.089.
(2) Lanka, P.; Deshpande, G. Combining Prospective Acquisition CorrEction (PACE) with Retrospective Correction to Reduce Motion Artifacts in Resting State FMRI Data. Brain Behav2019, 9 (8). https://doi.org/10.1002/brb3.1341.