Effective Preprocessing Procedures Virtually Eliminate Distance-Dependent Motion Artifacts in Resting State FMRI
Figure 7
Annotated processing flowchart for RS-FMRI analyses. If nuisance regressors are obtained before bandpassing and are to be projected out of the data after it is bandpassed, they must be bandpassed by the same filter before the projection. With too much censoring, one may end up with more regressors than data samples, and the preferred regression approach fails. Bandpass filtering censored or concatenated time series without taking into account temporal discontinuities is not recommended. The WMeLOCAL regressor is recommended, particularly for subject cohorts expected to exhibit high levels of head motion; it has the additional benefit of removing hardware artifacts that are hard to detect visually in the time courses of imaged volumes [3]. Slice-based physiological noise models [11] are projected from the data immediately after the despiking step because they are a function of slice acquisition time. A sample pipeline generating command is shown in example 5C at http://afni.nimh.nih.gov/pub/dist/doc/program_help/afni_proc.py.html).
