Neural barcoding representing cortical spatiotemporal dynamics based on continuous-time Markov chains

Populations of neurons form assemblies at many scales and display recurring spatiotemporal patterns of activity. In the cerebral cortex, these patterns of activity involve coordinated activity spanning large distances and anatomical regions subserving distinct functions. The constraints governing how these activity motifs transition over time is not known because conventional computational modeling and analyses collapse either the spatial or the temporal properties of the dynamics. Here, we use a continuous-time Markov chain (CTMC) modeling framework to probabilistically describe the temporal sequences elicited in large-scale complex cortical activity recorded with mesoscale imaging. This reveals a conserved dynamical structure across animals, with modular transitions serving as pseudo-"absorbing states." The parameters of the CTMC model are readily analyzed and used as a "neural barcode," a low-dimensional description of neural dynamics that is sensitive to cortical imaging applications, including pathological brain dynamics. This neural barcode provides a powerful computational tool to characterize cortical dynamics.