Analysis of fMRI data (or PET data) is almost never performed without first preprocessing the raw data. These preprocessing steps variably include temporal filtering to remove signal jitter across adjacent scans, removal of linear trends in signal intensity, spatial filtering (also called spatial smoothing), filtering to remove sources of periodic signal fluctuation related to vascular pulsation or breathing, intrasubject spatial alignment to remove movement across scans, and coregistration of BOLD data to the subject's structural MR1 to allow visualization of the images. Finally, intersubject alignment and warping (resizing) to a common stereotactic space are frequently performed. This final stage allows group statistical analyses on a voxelwise basis, but comes at the cost of spatial resolution and an understanding of individual variability (Brett, Johnsrude, & Owen, 2002).
Was this article helpful?