Inferring brain pathways of dynamic functional connectivity by diffusion and influence modeling
MetadataShow full item record
Growing research evidence from the functional brain imaging field indicates that the propagation of functional interactions provides an essential implementation of various brain functions. However, computational modeling of such pathways inferred from dynamic functional connectivity has been challenging, and there has been a lack of systematic investigation on the formation of transitions. In this work, we proposed a multi-stage functional brain pathway inference framework, by first modeling the dynamics of functional connectivity through sliding time window approach, followed by performing change point detection algorithm on the obtained connectivity strength dynamics. Then the causal relationships between brain regions/networks would be analyzed by a diffusion network inferring method (NETINF). NETINF would take the output of change point detection results, which are cascades of change point timestamps, as inputs, and then traced the most possible paths of diffusion and influence to infer the underlying pathways of change point propagations. We have applied the proposed framework on both the simulation data for validation and the task fMRI (tfMRI) dataset from the publically available human connectome project (HCP) Q1 release. The results show that by using the proposed model, we could obtain a set of consistent and neuroscientifically meaningful pathways from the tfMRI dataset.