I would like to calculate KCC using a larger neighborhood on every voxel. The idea for this is to compare it to a similiar method that looks at local connectivity using graph theory. In this paper, Sepulcre et al., (2010) used a radius of 14mm to assess neighboring voxels that had a similar time series correlation to the reference voxel, only excluding those with r < .25. I calculated that using a radius of 14mm is equivalent to 437 voxels (3x3x3mm in size). So, instead of using 27 neighboring voxels, is it possible to look at 437 or close?
Thanks!