Hierarchical Bayesian inference for community detection and connectivity of functional brain networks

• DOI: 10.48550/arxiv.2301.07386
• Publication Date: 2023-01-01
• Authors: Lingbin Bian; Nizhuan Wang; Leonardo Novelli; Jonathan M. Keith; Adeel Razi
• Keywords: Human Connectome Project; Generative model

Many functional magnetic resonance imaging (fMRI) studies rely on estimates of hierarchically organised brain networks whose segregation and integration reflect the dynamic transitions latent cognitive states. However, most existing methods for estimating community structure from both individual group-level analysis neglect variability between subjects lack validation. In this paper, we develop a new multilayer detection method based Bayesian block modelling. The can robustly detect weighted that give rise to hidden states with an unknown number communities retain networks. For validation, propose structure-based multivariate Gaussian generative model convolved haemodynamic response function simulate synthetic fMRI signal. Our result shows inferred memberships using hierarchical are consistent predefined node labels in model. is also tested real working memory task-fMRI data 100 unrelated healthy Human Connectome Project. results show distinctive structures subtle connectivity patterns 2-back, 0-back, fixation conditions, which may behavioural under task conditions.