This is a part of my Capstone Research. 

NSI is first proposed in my paper: An Improved Acquaintance Immunization Strategy for Complex Network

NSI can relect the regional influence and connectivity of nodes in a local network. When NSI used in medical image network, it is also effective.

1. To nodes varied in size along a line, a few nodes of small size can be compensated by its neighbour.

2. To nodes of large degree, it amplifies the original size of nodes, easier for identification of branch splits.
   

nodes marked with (original radius size, 1st neighbour NSI, 2nd neighbour NSI)

See closely with the variation of number between neighour nodes after using NSI

When comparing two images of different details, node degree will not intefere with the branch split identification. When using NSI as the standard, branch split is easily identified. Red nodes in above two images are nodes with large NSI, we can find the reddest node in a certain region is ideally paired (its neighbour nodes are all not so red as the split node).