Graduate Student Seminar: Brain Community Detection in Adults and Infant

The human brain comprises interacting regions that function both during rest and activity, with identifying these interactions being a key focus in neuroscience. While algorithms have mapped brain communities in adults and are increasingly applied to developmental studies, few have addressed determining the optimal number of communities in real-world datasets. This study benchmarks the Weighted Stochastic Block Model (WSBM) algorithm to estimate the optimal number of brain communities in infants. Using fMRI data from the Baby Connectome Project (BCP), the WSBM algorithm identified communities based on bootstrapped log-likelihood differences and a consensus algorithm. Our findings suggest that the infant brain comprises 15 communities and exhibits a non-assortative structure, aligning with prior research indicating incomplete development of higher-order association networks in infancy. These results provide insights into the early functional organization of the brain, supporting ongoing efforts to understand neural development.