Processing block: blur

In this step each volume is blurred with a 4 mm FWHM (full width at half max)
Gaussian kernel.

Background:

   There are 2 main reasons for blurring.  One is to reduce noise, by averaging
   across neighbors (with hopefully the same noise structure).  The other is to
   slightly expand statistically significant regions, so that there is better
   overlap of results in group space.  Since there is often reasonably large
   anatomical and functional variance across subjects (for example, there is
   typically ~10 mm of variability at the edge of the brain), 2 subjects might
   not overlap (well, if at all) in an activation area, even though it is the
   "same" area in some sense.

   There may not be a great gain in a large blur.  If the activation regions
   are fairly large, they should overlap fairly to begin with, especially with
   a small blur.  If the regions are small, a large blur could actually destroy
   the pattern (since non-active voxels are blurred into the active ones).

   So regardless of the size of activation, it seems to make sense not to blur
   too much.  We might suggest 0-8 mm, depending on many things (6 is common,
   4 is the default, and some are starting to use 0 or non-gaussian blurs).

   Also, the more blur in the data, the larger clusters will need to be when
   using AlphaSim and clustering for a multiple-comparison correction.

   Also, keep in mind that there is typically some blur in the data right out
   of the scanner, depending on the scanner.

The proc.FT script:

      # blur each volume of each run
      foreach run ( $runs )
          3dmerge -1blur_fwhm 4.0 -doall -prefix pb03.$subj.r$run.blur \
                  pb02.$subj.r$run.volreg+orig
      end

   For each run, blur each volume by a 4.0 mm FWHM gaussian kernel.  Options:

        -1blur_fwhm 4.0         : 4.0 mm, FWHM kernel
        -doall                  : blur all sub-bricks
        -prefix pb03...         : specify output dataset prefix
        pb02...volreg           : the input dataset (last item on line)


View the results in the afni GUI:

   At this point, controller [A] (the windows across the top of the screen)
   should be showing dataset pb02.FT.r01.volreg+orig and controller [B] should
   be showing pb01.FT.r01.tshift.  The coordinates should be (19.3 78 -5.7).

   So switch the underlay of controller [B] on the bottom to the blur output.

        - [B] : switch underlay to pb03.FT.r01.blur

   Besides the image being blurry, take a look at the time series.  What jumps
   out a bit that each of the times series looks less noisy than before (in the
   volreg dataset of controller [A], above).  The BOLD responses in the blur
   dataset are more clear than those in the volreg dataset.