Whole-cortex simulation reveals spatiotemporal patterns emerging from the interplay of network connectivity and intracellular dynamics

Abstract

Recent advances in Graphics Processing Unit (GPU) computing have allowed for computational models of whole-brain activity at unprecedented scales. In this work, we use desktop computers to build and simulate a whole-cortex mouse brain model using Hodgkin-Huxley type models for all the most active neurons in the mouse cortex. We compare the model dynamics over different types of connectivity, ranging from uniform random to realistic connectivity derived from experimental data on cell positions and the Allen Brain Atlas. By changing the external drive and coupling strength of neurons in the network, we can produce a wide range of oscillations in the gamma through delta bands. While the global mean-field behaviors of different connectivities share some similarities, an experimentally determined hierarchical connectivity allows for complex, heterogeneous behaviors typically seen in EEG recordings that are not observed in networks with nearest neighbors or uniform coupling. Moreover, our simulations reveal a wide range of spatiotemporal patterns, such as rotational or planar traveling waves, that are observed in experiments. Different traveling waves are observed with different connectivity and coupling strengths on the same connectivity. Our simulations show that many cortical behaviors emerge at scale with the full complexity of the network structure and ionic dynamics. We also provide a computational framework to explore these cortex- wide behaviors further.