Researchers identify new neural behavior pattern

This image was taken from Professor Jack Cowan’s video simulations of brain activity. The pattern shows how neurons fire in clustered bursts of activity instead of in spirals or waves. Cowan, who shares the discovery of this newly identified pattern with Tanya Baker, a Ph.D. student in Physics, describes the pattern as resembling the movements of flaming jellyfish.

University scientists, using forest fire dynamics to better understand brain activity, have identified a new pattern of simulated neural behavior that lends fresh insight into the brain’s inner workings. In this pattern, the neurons fire in clustered bursts of activity instead of in spirals or waves.

“A big debate has been going on for at least a decade now about how to interpret the firing patterns in the cortex,” said Jack Cowan, Professor in Mathematics, Neurology and the Committee on Computational Neuroscience. Although the matter remains unsettled, “This adds a new wrinkle to that debate.”

The new pattern, identified by Cowan and Tanya Baker, a Ph.D. student in Physics, resembles the movements of flaming jellyfish.

“Instead of a fire that spreads out, it stays localized, but it moves,” Cowan said. “You have a lot of little fires moving around, sensing each other and avoiding each other. Sometimes they collide and annihilate each other, and the whole thing looks like jellyfish moving around.”

In future work, Cowan hopes to link the simulations to the firing patterns of real neurons in the laboratory. “This is very much a work in progress,” he said. Meanwhile, he reported on his latest findings at the annual meeting of the American Association for the Advancement of Science last month.

The simulated behavior, if it occurs in real neurons, cannot yet be observed because it occurs at scales measuring the diameter of a few human hairs. This is too small to safely image with non-invasive medical imaging techniques.

The research was inspired by complexity theory, which shows a statistical relationship between phenomena as varied as forest fires, earthquakes, traffic-flow patterns and the size of cities. Complexity theory may also apply to brain-wave fluctuations, Cowan said.

The work of the late Per Bak, a Danish theoretical physicist, inspired Cowan to apply forest fire dynamics to brain-wave simulations. In 1990, Bak and two other scientists noted that widely varied phenomena could be described by power laws. These laws maintain identical proportionality between phenomena, regardless of the scale under consideration. For example, when the energy of earthquakes doubles, they become four times as rare. Similarly, for any given city size in the United States, there are four times as many cities that are half as big.

Bak proposed a concept called self-organized criticality to explain this phenomenon. He said these systems maintain themselves at the edge of chaos, like a slowly accumulating pile of sand that periodically collapses in an avalanche.

Cowan laid the foundation for his current studies 30 years ago, when he and Hugh Wilson, now of York University in Toronto, Canada, worked out a basic theory of how large-scale brain activity works. “Unfortunately, that theory doesn’t tell us too much about how different neurons in a population are correlated,” Cowan said.

A few years later, Cowan introduced to the scientific community his three-state neural model for how the brain works. In this model, Cowan described how neurons exist in one of three states: building up to a pulse, pulsing, and recovering from a pulse. When Bak’s forest-fire model came along in 1990, Cowan was struck by its similarity to his neural model.

“It’s clear that this forest-fire model is a version of the three-state model,” Cowan said. “Right away it triggered my interest.”

In the forest fire analogy, green trees correspond to neurons that are preparing to produce a pulse and “fire,” burning trees correspond to firing neurons, and burned trees correspond to neurons that are recovering from firing. In Bak’s forest fire analogy, fire would spread through a dense forest of green trees in a spiral pattern. This pattern is characteristic of densely connected neural networks, Cowan said.

The primary difference between the forest-fire model and neural networks is that the latter have inhibitory neurons that tend to dampen brain activity. The forest-fire model would be a more realistic neural model if it had a sprinkler system to simulate the action of inhibitory neurons, he said.

“You can build a forest-fire model with sprinklers that tend to stop the fire,” he said. “You have a few lightening strikes, but they don’t propagate much because the conditions for igniting the trees is much more difficult.”

To view short videos of Cowan’s simulations, visit: http://www-news.uchicago.edu/releases/videos/compneurosci/.