1) In the ppi_wrapper.m, how can I select *subject-specific* VOIs instead
of specifying one second-level VOI?

You either need a subject-specific nifti-file of the VOI OR you need the
VOI.mat used previously for this subject. The latter will create the
subject specific VOI nifti file and extract the data.
An alternative, would be to create an NxM matrix from the VOI.mat files,
where N is the number of timepoints per run and M is the number of
sessions). This would pull the BOLD data directly from the P.VOI field. The
NxM matrix would be put into the P.VOI field before running PPPI.m.

2) Is it possible to skip the VOI creation in gPPI?

You have to specify a VOI; however, you can directly feed in the timeseries
by entering the NxM matrix into the P.VOI field. This option was added on
Friday April 11th. Please download the newest version of the toolbox.
However, if you want to change the adjustment, then you will need to create
the VOI. As mentioned above the VOI can generated based on the existing
VOI.mat files.

3) Previously I have created two VOIs (one for each experimental run), and
I would like to use both again, but I guess the possibility to put multiple
VOIs into the variable 'regionfile' is not meant for this purpose, but only
when I would like to use different brain regions as seed regions? So how
can I use two VOIs for every subject?

You can only use one VOI.mat file per subject. To me, it makes no sense to
use different voxels for each experimental run. If the VOI.mat file has the
same voxels, then you have two options:
(a) Use the VOI.mat to create an NxM matrix and make this the value of P.VOI
(b) Select one VOI.mat file and list it as P.VOI, then the program will use
the same voxels for both runs.

If you really want to use different voxels for each run, then you will need
to Use the VOI.mat to create an NxM matrix and make this the value of P.VOI.

-----------
Another user emailed asking how to make subject specific VOI near the group
local maxima. At this time, gPPI can't find the closest maxima and draw a
sphere around this point. This could be coded up as part of the wrapper
though. However, it is possible to find a cluster within the group cluster
and limit the cluster to a specific size. To do this, the following
settings should be used:
P.VOI.VOI='group_resultmask.img';
P.VOI.masks={'spmT_0001.img' 'spmT_0002.img'};
P.VOI.thresh=[1.65 1.65];
P.VOImin=10;
P.VOI.exact=1;


*This option requires you to download the peak_nii toolbox.

If you just want to constrain the mask to be within the subject and group
clusters, then you can use the following settings:
P.VOI.VOI='group_resultmask.img';
P.VOI.masks={'spmT_0001.img' 'spmT_0002.img'};
P.VOI.thresh=[1.65 1.65];
P.VOImin=10;
P.VOI.exact=0;

P.VOI.VOI --> name of the ROI mask file to use as a starting point
P.VOI.masks --> a cell array of statistical images that you want to
use to constrain the mask (this could be subject specific t-maps)
P.VOI.thresh --> matrix of thresholds correpsonfing to each image in
P.VOI.masks

P.VOImin --> sets the minimum cluster size allowed. If the subject has
fewer, then it skips that subject/session.

If anyone has code for quickly finding the closest peak to the group peak,
then I could write a wrapper for generating subject specific spheres.