Activated cliques: Network-based activation detection in task-based FMRI

Human brain function has been widely believed as a network behavior. However, most previous activation detection methods in the task-based fMRI field were voxel-based, instead of network-based. For instance, the general linear model (GLM) has been widely used to fit the external stimulus curve with the fMRI BOLD signal of each voxel. In this paper, we present a novel network-based activation detection method to fit network-level measurement of the brain´s response with the external stimulus curve. The basic idea here is that based on the structural connectome constructed from DTI data, the propensity for synchronization (PFS) of combinations of three-nodes complete graphs, or cliques, is calculated from task-based fMRI signals, and the general linear model is then used to detect activations of the network-centric PFS curve. Further, given the intrinsically-established correspondences of structural connectomes and the derived complete graph cliques, the individual activation detection results are assessed across a population using the existing FSL FLAME framework to determine group-wise activated cliques during task performance. Our experimental results demonstrated that the network-based activation detection method is complementary to the widely-used voxel-based activation detection methods.