Comparative Evaluation of Assumption Lean Community Detection Methods for Human Connectome Networks

Abstract

Community detection on resting-state functional connectivity provides a principled lens on mesoscale organization in functional brain networks. Currently, the choice of community count K, lacks a standardized or principled guideline. We conducted a systematic benchmark of three assumption-lean approaches on weighted functional connectivity matrices: the Weighted Stochastic Block Model, Spectral Clustering, and K-means Clustering. Performance was assessed on synthetic networks with known ground truth and on three neuroimaging cohorts which included adult and infant datasets. We compared several strategies for selecting K, including the silhouette index and other approaches commonly used in the existing literature. These were evaluated alongside a likelihood-based criterion for the weighted stochastic block model that employs bootstrap confidence intervals for differences in log-likelihood between successive values of K. For the synthetic networks, WSBM and Spectral Clustering correctly identified the true number of communities, whereas K-means Clustering did not. In adult datasets, most indices did not yield a unique optimum, whereas the likelihood-based criterion selected K = 11, which is consistent with established sensory and association systems. In infants and toddlers, the same procedure supports a larger K around 15 and reveals developmentally distinct mesoscale architecture, including anterior and posterior subdivisions within default mode and fronto parietal systems.