Howdy-

3dNetCorr works by taking a time series data set ("-inset TTT") and a map of ROIs ("-in_rois RRR"). TTT and RRR must have the same grid. 3dNetCorr will calculate the average time series of each ROI in RRR, and store the correlation value (Pearson r or Fisher-transformed Z) in a correlation matrix.

The format of the ROI file RRR is as follows. A network is a set of ROIs in a single volume, and the number of ROIs in a network determines the size of the correlation matrix; a volume that has 10 ROIs (i.e., a network of 10 ROIs) will have a 10x10 correlation matrix in its output. Each ROI is defined as the set of voxels having the same integer value. So, if RRR contains information about a single network, it will have a single volume. If that network has 10 regions, then there will be, say, some voxels with values 1, some voxels with value 2, etc. up to 10. Essentially, each integer labels the ROI; note that the integer values do not have to be consecutive (1...N), but could be any positve integers.

If you want to look at multiple networks with a single call to 3dNetCorr, separate networks can be defined in separate volumes that are input. Thus, RRR could have multiple volumes, each one treated as a separate network and getting a separate correlation matrix output. In each volume, an ROI of the network is defined in the same way: just a set of voxels with the same integer value.

In your above example, if your created ROIs all have the same value and are not touching, then you could use 3dROIMaker to assign each a different integer value. Or, if your created ROIs all have different integer values after your 3dUndump command, then you should be good to go with 3dNetCorr.

--pt