Etay Ziv1,
Julia P. Owen1, Yi-Ou Li1, Eric J. Friedman2,
Pratik Mukherjee1
1University
of California, San Francisco, San Francisco, CA, United States; 2International
Computer Science Institute , Berkeley, CA, United States
We apply a Variational Bayesian approach to modularity analysis of the structural connectome in normal adult subjects and compare this VBMOD algorithm to the two most widely used module detection algorithms within the brain network community. We demonstrate the superiority of VBMOD to these existing methods in precision, accuracy and robustness to noise, both in module identification and cardinality inference. These findings are consistent over a broad range of thresholds used to binarize the networks. Our results are of interest to researchers in the connectomics and diffusion tractography literature.