One of the instructors at a recent AFNI bootcamp suggested that, for sagittal slices, 2D registration might be performed on the slices prior to 3D registration of the volumes. The suggestion was that nodding motion might be effectively corrected by the 2D registration. Does anyone have an idea for how to best account for this stepwise registration in the deconvolution? I could include 9 regressors (6 from 3D, plus 3 from 2D), but I imagine potential collinearity issues.
Thanks for your thoughts.
Tyler