Coarse-graining to investigate cerebral cortex dynamics: The case for experimental simplification of biological phenomena

Abstract

Advances in multi-channel/multi-detector recordings and data analysis over the last decades have led to an explosion in the exploration of complex neural dynamics in mammalian cortex. Powerful methods have been applied to investigate such dynamics, including connectivity measures (correlation, causality, resting state synchrony, etc.), spatiotemporal pattern analyses, and finite-element modelling based on model neurons. These methods were initially applied to data from simple experimental models such as invertebrate neurons/ganglia/tecta, cell cultures, and organotypic slice preparations. Advances in the field have triggered the expanded use of such measures on more complex data, for example to mammalian ex vivo preparations, anesthetized preparations, and mammalian awake behaving preparations. With the increasing surgical, behavioral, and physiological complexity of the preparations themselves, less invasive measurement methods such as optical recordings, massively implanted arrays, or fMRI and other electromagnetic methods must be used to ensure robustness; however, these measures tend to feature lower signal-to-noise ratios, and are often prone to various biases. Furthermore, the high dimensionality of the data itself leads directly to potential errors in programming of analysis algorithms and overinterpretation of statistically significant but biologically insignificant findings. Given this situation, we advocate for the complementary use of the classical biological approach: the use of simplified preparations which may be limited in scope, but which highlight fundamental principles. We illustrate this approach with three experimental examples which use experimental and observational approaches to coarse-grain dynamic spatiotemporal activity patterns, to make coarse-graining observations of clinically relevant oscillations, and to coarse-grain complex behavior in mammalian discrimination learning.