Making Multiple ROIs for 3dNetCorr ROI Analysis


I am new to both AFNI and Ubuntu(on Windows). I want to make 18 ROIs and extract various data on these 18 ROIs like correlation matrix, avergae time series etc.

I tried the following and it didnt work.

I saved a txt file with the xyz coordinates of all 18 roi.'s 1 in each line. I also added a number as a 4th dimension. Then used 3dUndump to make a mask but now when i run 3dNetCorr it gives me an error that it needs more that 1 sub brick.

How can i workaround this problem.


What was the 3dNetcorr command that you entered? An example one would be:

3dNetCorr                                  \
         -inset TIME_SERIES_DSET               \
         -in_rois ROI_MAP_DSET              \
         -mask MASK_DSET                     \
         -prefix OUTPUT_DSET

where there are two main inputs:

  • a TIME_SERIES_DSET, which is a 3D+time dataset of time series values throughout a volume,
  • a ROI_MAP_DSET, which is a map of ROIs (each ROI is a set of voxels with the same integer value, whether contiguous or not),
    and a secondary/optional input, the MASK_DSET (a mask to apply to further restrict the volume analyzed, if you want).

It sounds like you have made the ROI_MAP_DSET already from using 3dUndump (and you should use the AFNI GUI to verify that the spheres are in the places you want).

Where is your TIME_SERIES_DSET coming from?



Sorry for the late reply. I for some reason did not get a mail about this so assumed that noone answered. I am further along my quest but i am struggling with dealing with overlapping ROIs.

I have 18 ROI centers. I use 3dUndump to make 18 ROI masks of radius 10. Because of the large size, there are overlaps. So when i combine them using 3dcalc using formula (a+b+c+d+e+f+g+h+i+j+k+l+m+n+o+p+q+r) and then do 3dNetCorr i end up with 20 columns with codes 1-18 , 25 and 28. Can you suggest how to deal with this? I would still want 18 ROIs in the end.


HI, Prateek-

You have asked this question in 2 places; I will answer it on the thread you started specifically for it, here:,164043,164043#msg-164043