A Markovian neural barcode representing mesoscale cortical spatiotemporal dynamics.

Abstract

Mesoscale cortical dynamics consist of stereotyped patterns of recurring activity motifs, however the constraints and rules governing how these motifs assemble over time is not known. Here we propose a Continuous Time Markov Chain model that probabilistically describes the temporal sequence of activity motifs using Markov Elements derived using semi-binary non-negative matrix factorization. Although derived from a discovery sample, these can be applied to new recordings from new mice. Unwrapping the associated transition probability matrix creates a Markovian neural barcode describing the probability of Markov element transitions as a compact and interpretable representation of neocortical dynamics. We show broad utility across a range of common mesoscale cortical imaging applications, ranging from time-locked events to pathological models. Moreover, it allows the discovery of new and emergent Markov Elements that unmask the flexibility of constraints governing cortical dynamics. The Markovian neural barcode provides a novel and powerful tool to characterize cortical function.