There are now two new programs as part of AFNI for calculating
widely-used resting state functional connectivity (RSFC) parameters
for resting state fMRI (rs-fMRI). These are: A) 3dRSFC and B) 3dReHo.

Both programs should be:
+ simple to use,
+ fairly efficient time/memory-wise,
+ provide informative parameters for RSFC studies in a single (or
nearly so) swoop.

The programs were written by PA Taylor, and feedback in the form of
questions, complaints, suggestions for increased usefulness, etc. can
be directed to him either via email (neon.taylor _at_ gmail.com), or,
better yet, on the AFNI community board here. NB: actually, 3dRSFC is
written on top of the existing and quite useful 3dBandpass (written
previously by RW Cox), of which all well-tested functionality remains,
with some new outputs and options added to constitute 3dRSFC.


DESCRIPTIONS:

A) -----------------------------------------

3dRSFC is a new version of 3dBandpass (literally, written on top of
that program), and the former keeps all of the same standard
processing options of the latter for producing a standard resting
state, low-frequency fluctuation (LFF) time series (e.g., 0.01-0.1 Hz
or thereabouts). In addition, 3dRSFC allows one to simultaneously
calculate several standard/useful RSFC parameters across the brain and
output 3D maps of them. The parameters are: ALFF, fALFF, mALFF, RSFA,
fRSFA and mRSFA. For parameter derivations/references, see:
ALFF/mALFF -- Zang et al. (2007),
fALFF -- Zou et al. (2008),
RSFA -- Kannurpatti & Biswal (2008).
(and we introduce fRSFA and mRSFA, calculated in analogy to ALFF
counterparts).

Essentially, ALFF and RSFA are the L_1 and L_2 norms, respectively, of
the low-frequency bandpassed LFF time series; fALFF is just ALFF
scaled by the sum of amplitudes of the total, unbandapassed time
series (and analogously for RSFA with the square-rooted sum of power
terms); mALFF is ALFF scaled by the mean ALFF value in the brain (and
analogously for mRSFA). The amplitudes are calculated at appropriate
times during the `processing' (which is the point of calculating them
*during* the processing steps, listed just below for this program).

The `processing' is done by *exactly* the same means as 3dBandpass
would do. Therefore, one inputs raw 4D rs-fMRI data sets and selects
any of the following `processings' (from the help description):

(0) Check time series for initial transients [does not alter data]
(1) Despiking of each time series
(2) Removal of a constant+linear+quadratic trend in each time series
(3) Bandpass of data time series
(4) Bandpass of -ort time series, then detrending of data
with respect to the -ort time series
(5) Bandpass and de-orting of the -dsort dataset,
then detrending of the data with respect to -dsort
(6) Blurring inside the mask
(7) Local PV calculation
(8) L2 normalization,

and the output is a 4D data set: the `processed' rs-fMRI data set
(i.e., containing most likely just the LFFs). Along the way,
appropriate LFF amplitudes are calculated (and the trick being that
for fALFF, one needs amplitudes of *non*-filtered time series for the
denominator), and so then one gets 3D maps of RSFC parameters, as well
as a resting state time series. There are some options for selecting
which parameters are output and various things listed in the `3dRSFC'
help description.

3dRSFC summary:
+ input is raw resting state data.
+ outputs are:
- LFF and possibly smoothed, detrended, etc. time series (4D data set)
which one would get from 3dBandpass,
- parameter maps of ALFF/fALFF/RSFA/etc. (3D data sets).


B) --------------------------------

3dReHo returns 3D whole brain maps of ReHo (regional homogeneity),
whose values are just just a renaming of the Kendall's W (or Kendall's
coefficient of concordance, KCC, (Kendall & Babington Smith, 1939))
for a set of time series. Application to fMRI data was described by
Zang, et al. (2004), where it was applied to the study of both task
and resting state functional connectivity.

In this case, one would probably input a filtered/processed rs-fMRI
time series (i.e., one containing only LFFs, having been put through,
for example, 3dBandpass or the new 3dRSFC). The output would be a 3D
whole-brain map of ReHo values, and one can also get an additional 3D
map which has the Friedman chi-square value for each ReHo value, as
an option.

Technical notes about ReHo calculation: one can select whether to use
neighborhood size per voxel of 7, 19 or 27 (probably just 27 should be
used); at the edges, the neighborhood automatically shrinks in order
to not include non-brain null series; corrections for `ties' in the
ordering have been accounted for, though they typically make very small
differences.