HowTo #1:
			  Experiment Background

Block design - measurement of time delay to visual stimulus


The Experiment:

    In this example, subjects were presented with a checkered, rotating
    hemifield.  One run consisted of 30 seconds of rest, followed by 6
    pairs of 30 seconds ON (rotating hemifield) and 15 seconds OFF (rest)
    intervals, and lastly followed by 20 seconds of rest at the end of the
    run.  Thus each run was 320 seconds long, with a TR of 2 seconds, providing
    160 volume images.


The data:

    This EPI data consists of 160 volumes, 18 coronal slices each.  These
    slice images were taken of the back of the brain (28.8 mm posterior to
    96.8 mm posterior).  Each slice is a 64 by 64 voxel image, with a voxel
    resolution of 3.75 mm on each side.  The thickness of each of the 18
    slices is 4.0 mm.

    The SPGR anatomical data consists of 124 axial slices, from 50.6 mm
    inferior to 84.7 mm superior.  Each slice is a 256 by 256 voxel image,
    with each voxel being a 0.938 mm square.  The slice thickness (the
    distance between 2 slices) is 1.0 mm for the anatomical images.

    Note the difference in resolution between the SPGR data and the EPI
    data.  The EPI data has a much lower resolution.  A single voxel of
    the EPI data is basically 4 times the size of a single voxel of SPGR
    data, along each of the 3 axes.  A single voxel of EPI data will therefore
    occupy the same space as approximately 64 voxels of SPGR data.

    The SPGR data is located under the SPGR_data directory, in 124 I-files 
    (i.e., I.001 through I.124).

    The EPI data is under the EPI_data directory, in a sequence of directories
    generated by a GE scanner: 003, 023, 043, 063, 083 and 103.  This is
    slightly more than 2 runs of data contained in almost 6000 I-files.


Background:

    Because of the retinotopic organization of some visual cortical areas, 
    stimuli in different parts of the visual field stimulate different parts of
    the visual cortex. To determine the mapping between the visual field and 
    retinotopic areas, we use a visual stimulus that repeatedly sweeps through 
    the visual field in 30 seconds, every 45 seconds. This causes different 
    portions of retinotopic areas to be activated at different times depending 
    on the portion of the visual field that they represent. For example, if a 
    stimulus starts as a small annulus in the middle of the visual field 
    (foveal) and expands to the peripheral part of the visual field 10 seconds 
    later, then we expect cortical areas mapped to the peripheral part of the 
    visual field to respond with a 10 second delay relative to areas mapped to 
    the foveal part. Thus by estimating the response delay of the activated 
    voxels we can determine which part of the visual field they respond to. 

    We predict that different portions of the visual cortex will respond with 
    different delays depending on what portions of the visual field they 
    respond to.

    This is a simplified exposition of retinotopic analysis. For more details 
    see: [1-3].  Estimating the response delay of activated voxels is done 
    using the program 3ddelay (or the AFNI plugin Hilbert Delay 98). 
    The program implements a computationally efficient way of estimating the 
    response delay of each voxel along with its cross-correlation coefficient 
    which is used to determine whether a voxel is activated by the stimulus or 
    not. The response delay is estimated using the Hilbert Transform of the 
    cross correlation function. For nauseating details about FMRI response 
    delays and delay estimation algorithm see [4, 5].


References:

    1.	Engel, S.A., et al., fMRI of human visual cortex. Nature, 1994. 
        369(6481): p. 525.

    2.  Sereno, M.I., et al., Borders of multiple visual areas in humans 
        revealed by functional MRI. Science, 1995. 268(268): p. 889-893.

    3.	DeYoe, E.A., et al., Functional magnetic resonance imaging (FMRI) of 
        the human brain. J. Neuroscience Methods, 1994. 54: p. 171-187.

    4.	Saad, Z.S., E.A. DeYoe, and K.M. Ropella, Estimation of FMRI Response 
        Delays. Neuroimage, 2002: p. in press.

    5.	Saad, Z.S., et al., Analysis and use of FMRI response delays. Human
        Brain Mappign, 2001. 13(2): p. 74-93.

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

Steps taken to analyze the data:

    We have created and run a shell script (called @ARzs_analysis) which
    performs the following steps.

    o  Put both the SPGR data and the EPI data into AFNI bricks.

      - The SPGR data is all under the SPGR_data directory, so this can easily
        be put into an AFNI brick, using the 'to3d' program.

      - The EPI data is more confusing.  Although 'to3d' alone works well, we
        will use the 'Ifile' program to create a Unix script called 
        'GERT_Reco'.  Simply running this script will create the AFNI bricks 
	(the script will call 'to3d' with the relevant parameters).

    o  All of our EPI volumes (160 volumes in each of 2 runs) are 
       registered with the volume in run 2, time point 156 (sub-brick
       155).  This is done to correct for small head movements during the
       scanning session.  The program '3dvolreg' is used for this step.

    o  The two EPI run datasets are averaged using '3dcalc'.

    o  A reference file was created using the program 'waver'.  This file
       represents an ideal response function over the course of the
       run.  This file will be needed in the next step as an input to the
       program '3ddelay'.

    o  Here we actually create a functional brick.  The program '3ddelay'
       is used to compute, among other things, a delay sub-brick and a
       correlation sub-brick.  The delay sub-brick contains the data we
       are actually after (the time lag between the ideal response
       function and the measured response function at each voxel).  The
       correlation sub-brick shows the correlation between the measured
       data for each voxel and the ideal time series.  This might suggest
       those voxels for which the delay is relevant.