Low-Dimensional Spatiotemporal Dynamics Underlie Cortex-wide Neural Activity

Camden J Macdowell,Timothy J Buschman

doi:10.1016/j.cub.2020.04.090

Abstract

Cognition arises from the dynamic flow of neural activity through the brain. To capture these dynamics, we used mesoscale calcium imaging to record neural activity across the dorsal cortex of awake mice. We found that the large majority of variance in cortex-wide activity (∼75%) could be explained by a limited set of ∼14 "motifs" of neural activity. Each motif captured a unique spatiotemporal pattern of neural activity across the cortex. These motifs generalized across animals and were seen in multiple behavioral environments. Motif expression differed across behavioral states, and specific motifs were engaged by sensory processing, suggesting the motifs reflect core cortical computations. Together, our results show that cortex-wide neural activity is highly dynamic but that these dynamics are restricted to a low-dimensional set of motifs, potentially allowing for efficient control of behavior.

Highlights

The brain is a complex, interconnected network of neurons
Our results suggest cortex-wide neural activity is highly dynamic but that these dynamics are low-dimensional: they are constrained to a small set of possible spatiotemporal patterns
A translucent-skull prep provided optical access to dorsal cortex, allowing us to track the dynamic evolution of neural activity across multiple brain regions, including visual, somatosensory, retrosplenial, parietal, and motor cortex (Figure 1A, inset and Figure S1A [17])

Summary

Introduction

Neural activity flows through this network, carrying and transforming information to support behavior. Previous work has associated particular computations with specific spatiotemporal patterns of neural activity across the brain [1,2,3]. Specific spatiotemporal patterns of cortical regions are engaged during goal-directed behaviors [7], motor learning and planning [8, 9], evidence accumulation [10], and sensory processing [11]. Previous work has begun to codify these dynamics, either in the synchronous activation of brain regions [3, 12] or in the propagation of waves of neural activity within and across cortical regions [13, 14]. This work suggests cortical activity is highly dynamic, evolving over both time and space, and that these dynamics play a computational role in cognition [15, 16]

Objectives

Methods

Results

Discussion

Conclusion