REGIONAL METASTABILITY OF WHOLE-BRAIN KURAMOTO NETWORKS ACROSS MULTIPLE CONNECTOME SCALES

“Metastability” expresses the ability of a dynamical system to transition between multiple quasi-stable attractors. Metastability in the brain is thought to be related to the brain’s ability to integrate functional components and facilitate cognitive flexibility. beim Graben et al, Front. Comput. Neurosci., 2019 combined structure-function MR data to propose a brain hierarchical atlas (BHA) and to construct a state space partitioning that is maximally metastable. They identified a partitioning level with 40 structural modules (network nodes) where metastability peaked. Whether the same partitioning applies to neuroimaging data at a different temporal and spatial scales is unknown. This could impact on understanding brain dynamics in general, and on studies fusing data at multiple spatio-temporal scales. We used the BHA approach to construct whole brain networks at different structural partitioning resolutions, and a phenomenological neural dynamic model of Kuramoto oscillators to simulate their dynamics. At each spatial level, we obtained the level’s maximum metastability by optimising two scalar parameters: the global coupling strength and the mean conduction delay of the connections between nodes. Given the oscillatory nature of the Kuramoto networks, we defined metastability in terms of Spectral Entropy of the nodes. We demonstrate the three-way relationship between spatial scale, model parameters and network metastability (averaged metastability over the nodes). We discuss the distribution of the regional (node level) metastability over the spatial scales. We highlight the importance of the methods and features used for structural partitioning on the estimate and distribution of metastability, and its impact on multimodal brain imaging studies.