I’m confused about the aspect of motion correction in the fMRIprep pipeline. FMRIprep concatenates the transform from susceptibility distortion correction, co-registration and head-motion estimation to map the EPI images to standard space. According to what I understand, this means that head-motion correction is performed.
Then why is it outputing head motion parameters as coufounds as well ?
Should we do an additional motion correction by regressing those head motion related confounds ?
Thanks for your help,
fMRIPrep realigns the data array in its final interpolation space (that puts the EPI in standard space)
However, head-motion parameters (and derivations, such as Framewise Displacement) can be used in modeling as nuisance regressors. For instance, after sudden and acute motion spikes there’s typically a spin de-magnetization (usually visible as a dip in the overall global signal) that you will definitely not capture with the realignment and these parameters can help model that.
What regressors should be included is a decision of the researcher. This section of the documentation - https://fmriprep.org/en/stable/outputs.html#confounds is pretty good in attempting to understand this.
Thanks @oesteban, that’s clearer now.