Check on/review a few things here:

1. recall what input datasets were passed to the afni_proc.py
2. note what files are populating the results directory
3. look at the original input data in afni, to recall points of interest
4. close afni

---------------------------------------------------------------------------

1. Recall that all of the input came from the FT directory, including the 2
   stimulus timing files, an anatomical dataset and 3 EPI datasets.

   % ls FT

      AV1_vis.txt, AV2_aud.txt     : stimulus timing files
      FT_anat+orig                 : anatomical dataset
      FT_epi_r1+orig               : EPI run 1
      FT_epi_r2+orig               : EPI run 2
      FT_epi_r3+orig               : EPI run 3

   Recall also that each AFNI dataset consists of 2 files, a .HEAD text file
   that describes the data (location, grid, orientation, description, etc.),
   and a .BRIK file which contains the raw data.


2. Note what files are populating the "results" directory.

   % ls FT.results

   Files will be added to this directory as processing proceeds.  The very last
   files to be written are blur_est.FT.1D (with residual blur estimations) and
   @epi_review.FT (a script to quickly review the original data).

   Most importantly, note the format of the pre-processed EPI datasets.  For
   example, consider the .BRIK files for only run 1:

        pb00.FT.r01.tcat+orig.BRIK
        pb01.FT.r01.tshift+orig.BRIK
        pb02.FT.r01.volreg+orig.BRIK
        pb03.FT.r01.blur+orig.BRIK
        pb04.FT.r01.scale+orig.BRIK

   These are named starting with 'pbNN', where NN is the index of the given
   processing block.  So 'tcat' is processing block 00, 'tshift' is block 01,
   then 'volreg', 'blur' and finally 'scale', the last pre-processing step
   before regression.  The center part of the name is the subject ID, followed
   by the run number.  For example, consider pb02.FT.r01.volreg:

        pb02    : processing block index 02
        FT      : subject ID
        r01     : run #1
        volreg  : pb02 is the volume registration processing block

   These names were chosen to be descriptive and alphabetical (in terms of
   processing order).


3. Start afni in the FT directory and take a quick look at the original data.

   % afni FT

   A. Open up each image window (axial, sagittal, coronal), set the underlay
      to be FT_anat and the overlay to be FT_epi_r1. and make a few notes:

        - The alignment is not perfect, but it is pretty good.  Recall that CSF
          makes the EPI brightest.  So the red parts of the EPI should overlay
          the ventricles, etc.

        - The coverage of the EPI is much smaller than that of the anat, though
          it covers most of the brain (but only part of the cerebellum).

        - There is not much EPI signal dropout within the brain.

   B. Turn off the overlay and choose FT_epi_r1 as the under lay.
      Open an axial graph window.
      Jump to xyz coordinates "19.3 78 -5.7", to be sure of what is seen.
      Close the Coronal window to save screen space.
      In the graph window, type an 'A' for auto-scaling (or Opt->Scale->AUTO).

      To jump to those xyz coordinates:

        - put the mouse in an image window
        - right-click and select "Jump to (xyz)"
        - type (or copy-and-paste) 19.3 78 -5.7 as xyz coordinates
        - hit Set

      A few things are apparent when looking at the Axial Graph window:

        - Some anatomical detail can be seen, even though this is EPI data.

        - Typical values seem to be between 1000 and 2000.

        - The first first TRs are very high in magnitude.  So those are the
          pre-steady state TRs which will be removed in the processing script.

        - There is a spike around index 44.  Moving to that index (using the
          '>' key in the graph window) and watching the sagittal image, it
          becomes clear that is due to a subject head motion.  Nodding motions
          are the most common.

          The motion should be taken care of by 3dvolreg and by using the
          motion parameters as regressors of no interest.

          ** A large motion like this should really be censored.  Note the
             "-regress_censor_motion 0.3" option applied in the more advanced
             afni_proc.py script, s02.ap.align .

        - This voxel shows a clear up and down pattern that we expect means
          there was a response to our stimuli.

4. Close afni.  We will run it again from the results directory.

   Advanced: Of course it is not necessary to close afni.  One could 'read' in 
             the new results directory (perhaps once processing has finished)
             and then 'switch' to it, or even tell afni to read both the FT
             and FT.results directories to begin with.