I am using censor motion in first-level analysis, and I have two questions: The first one is whether I should censoring only the time points where the enorm exceeds a certain threshold, or should I delete the time points before and after as well, or delete the time point itself and the previous one?
The second question is regarding my task-fMRI where a run has 333 time points. When a certain percentage of time points are deleted, at what percentage should I discard the entire run? I couldn't find any recommendations for this in the literature.
When you use afni_proc.py and include motion censoring with -regress_censor_motion 0.3, for example, the proc script calculates the Enorm (Euclidean norm) across the time series, which is the square root of the sum of squares of motion parameters for each point. Note that this uses the "first difference" between successive time points. So, when afni_proc.py censors where "Enorm > 0.3", it censors both time points whose difference was suprathreshold.
Note that we typically also add a separate criteria for censoring volumes based on where the fraction of outliers within the mask is quite high; this is done in AP with, like, -regress_censor_outliers 0.05, which would censor individual volumes where more than 5% of the time voxels have an outlier at that time point. This censors just that single volume.
There is more description of this in this AFNI Academy video/playlist of motion correction and alignment, and in these papers:
Reynolds RC, Glen DR, Chen G, Saad ZS, Cox RW, Taylor PA (2024). Processing, evaluating and understanding FMRI data with afni_proc.py. Imaging Neuroscience 2:1-52. https://doi.org/10.1162/imag_a_00347
Taylor PA, Chen G, Glen DR, Rajendra JK, Reynolds RC, Cox RW (2018). FMRI processing with AFNI: Some comments and corrections on ‘Exploring the Impact of Analysis Software on Task fMRI Results’. bioRxiv 308643; doi:10.1101/308643 https://www.biorxiv.org/content/10.1101/308643v1.abstract
As to know at what point the censoring has become "too much" to use that run of data, that is really an open question. That should be part of a pre-study plan, but can be tricky because some participant populations are inherently more prone to motion than others, and one might consider different allowances. How to deal with motion on this scale was discussed a lot in different FMRI Open QC Project papers, which are assembled here. It is important to balance individual data quality with not biasing group results and also not wasting data. Our own discussions of this and considerations, as well as how the APQC HTML report warns about motion, are here:
Thank you for your suggestions. My MRI data only scanned a specific brain region with a very small FOV. Regarding the -regress_censor_outliers 0.05 you mentioned, I had considered it, but considering the small number of voxels in the scanned region, the proportion of outliers at thresholds like 0.05 or 0.1 differs significantly between the whole brain and a small region due to the difference in base size. Therefore, I decided to turn off this function. Is this change reasonable?
Ah, this is a good example of how every dataset has its own considerations for specifying processing. That makes sense about not including the outlier fraction.
Or, also, I guess you could set the fraction to be quite large, perhaps (maybe 0.5, so 50%), juuust to be able to keep the check there, and then you will see the outlier fraction plot usefully at the end. If there is a volume with 50% outliers, even for a small volume, you would likely still want to censor that, perhaps?
--pt
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.