|
Abstract:
|
Individual differences in neurodevelopment contribute to a broad range of psychiatric disorders. Regardless of the precise mechanism or behavior, the underlying assumption of all neurodevelopmental models of risk is that at the population level, there exist subgroups of individuals that share similar patterns of neural function and development and these subgroups reflect different risk profiles. However, the presence of multiple neurodevelopmental subgroups, as defined by brain functional connectivity has not been assessed previously. In this work, we propose a nonparametric Bayesian clustering scheme based on brain functional connectomes across both resting and task states, which dissects subtypes integrating multidimensional brain functional organizations. We simultaneously learn the network parcellations under each cognitive construct and identify the informative subnetwork units playing the key roles during subtyping. To facilitate posterior inference, we develop an efficient variational Bayes algorithm which allows the practical use of the proposed network clustering model. After intensive simulations, we apply the method to the motivated Adolescent Brain Cognitive Development
|
