Thanks for reply.
I am taking coupled voxel coordinates and creating co-activation map (as a nifti image) followed by decoding this coactivation map. I did so since I want two level of analysis. One for each coupled voxels and one for overall group. Logical it sounds me correct. However, any suggestion is welcome.