Dynamic Bayesian brain network partition and connectivity change point detection

Multiple recent neuroimaging studies revealed that functional interactions within brain regions are locally clustered into small sub-networks where different dynamics of functional interaction occur. However, integration models investigating such functional brain dynamics have been rarely explored. In this paper, a novel Bayesian inference model is developed to partition the brain regions into different sub networks and to simultaneously segment temporal sequence of each sub network into several quasi-stable blocks based on the interaction dynamics among regions. The proposed model has been evaluated and validated by two different simulation models. Also, the model has been applied to a working-memory task-based fMRI dataset and interesting results on both dynamic sub networks and change points were obtained.