summary: Sleep can be detected by patterns of brain activity lasting just milliseconds, according to a new study. The study shows that small areas of the brain “blink” momentarily into wakefulness or sleep, calling into question conventional views of sleep and wakefulness.
Using advanced neural network analysis, the researchers discovered high-frequency patterns that define sleep, and these findings may aid in the study of neurodevelopmental and neurodegenerative diseases associated with sleep disorders.
Key Facts:
- Millisecond detection: Sleep is identified by neural activity that lasts for a few milliseconds.
- Flicker: Small areas of the brain can briefly switch between sleep and wakefulness.
- Research impact: These findings may aid in understanding diseases associated with sleep disorders.
sauce: University of California, Santa Cruz
Sleep and wakefulness. These are completely different states of existence that define the boundaries of our daily lives. For years, scientists have been measuring the differences between these instinctive brain processes by observing brainwaves. Sleep is characterized by slow, prolonged waves measured in tenths of a second that travel throughout the organs.
For the first time, scientists have discovered that sleep can be detected by neural activity patterns that last just a few milliseconds – a thousandth of a second – revealing a new way to study and understand the fundamental brainwave patterns that govern consciousness.
It has also been shown that small areas of the brain can briefly “flicker” awake while the rest of the brain remains asleep, and vice versa.
These findings are described in a new study published in the journal Nature Neuroscienceis a collaboration between the laboratories of Keith Hengen, assistant professor of biology at Washington University in St. Louis, and David Haussler, distinguished professor of biomolecular engineering at the University of California, Santa Cruz. The research was conducted by doctoral students David Parks (University of California, Santa Cruz) and Aidan Schneider (Washington University).
Over four years of research, Parks and Schneider trained neural networks to study patterns within vast amounts of EEG data, discovering patterns that occurred with unprecedented frequency and called into question long-held fundamental concepts about the neurological basis of sleep and wakefulness.
“There is much to be gained by using powerful tools and new computational methods to question our most basic assumptions and rethink the question, ‘what is a nation?'” Hengen said.
“Sleep and wakefulness are the biggest determinants of behavior, and everything else falls off of that. So if we don’t understand what sleep and wakefulness actually are, we’re missing out.”
“For us as scientists, it was a surprise to discover that another part of the brain actually takes a little nap while the other part is awake. Many people may have already suspected that this phenomenon occurs in their spouses, but the surprise may be that there is no gender bias,” Hausler joked.
Understanding Sleep
Neuroscientists study the brain through recordings of electrical signals of brain activity, known as electrophysiological data, and see voltage waves that rise and fall at different paces, mixed with the spiking patterns of individual neurons.
The researchers worked with data from mice at the Hengen Institute in St. Louis, outfitting the freely moving animals with extremely lightweight headsets that recorded brain activity in 10 different brain regions for months at a time, tracking voltages from small populations of neurons with microsecond precision.
Feeding this amount of data created petabytes (a million times gigabytes) of data. David Parks led the effort to feed this raw data into artificial neural networks that can find incredibly complex patterns, distinguish between sleep and awake data, and find patterns that human observation may miss.
Collaboration with a shared academic computing infrastructure at the University of California, San Diego, enabled the team to work with large amounts of data on the scale used by large companies such as Google and Facebook.
Knowing that sleep is traditionally defined by slow-moving waves, Parks began feeding the neural network smaller and smaller chunks of data, instructing it to predict whether the brain was asleep or awake.
The researchers found that their model could distinguish between sleep and wakefulness from just a few milliseconds of brain activity data, which was a shock to the team, as it showed that the model couldn’t have relied on slow-moving waves to learn the difference between sleep and wakefulness.
Just as you can’t tell if a rhythm is slow by listening to a thousandth of a second of a piece of music, it’s impossible for a model to learn a rhythm that occurs over several seconds by looking at only a few randomly separated milliseconds of information.
“We’re looking at information at a level of detail never before seen,” Hausler said. “Previously, the assumption was that you wouldn’t find anything there, that all the relevant information would be in the lower frequency waves.
“The paper shows that ignoring conventional measurements and looking only at the millisecond-long high-frequency measurement details is enough to determine whether tissue is asleep. This suggests that something is happening on a very fast timescale, providing new clues about what happens during sleep.”
Meanwhile, Hengen was convinced that Parks and Schneider’s findings were so at odds with basic concepts drilled into them by years of neuroscience education that they must have missed something, and he challenged Parks to provide more and more evidence that the phenomenon might be real.
“This experience forced me to ask myself, ‘To what extent are my beliefs based on evidence, and what evidence do I need to overturn them?'” Hengen said.
“It was really a game of cat and mouse because I told David, [Parks] “He would present me with evidence over and over again, trying to prove things, and then he’d come back and say, ‘Look at this!’ It was a really interesting process for me as a scientist to have my students take down these towers brick by brick and then have to accept that.”
Local Patterns
Because artificial neural networks are essentially black boxes and don’t report what they’ve learned, Parks began peeling back layers of temporal and spatial information to try to understand what patterns the models were learning.
Eventually, they got to the point where they were looking at chunks of brain data just one millisecond long, and the highest frequencies of brain voltage fluctuations.
“We took all the information that neuroscience has used over the last century to understand, define and analyze sleep and asked, ‘Can our models learn under these conditions?'” Parks says. “This allowed us to look at signals that we couldn’t understand before.”
Examining these data allowed them to determine that ultrafast patterns of activity among just a few neurons are the building blocks of sleep that the model detects – and importantly, these patterns cannot be explained by traditional slow, widespread waves.
The researchers hypothesized that the slower-moving waves might help coordinate faster, more localized patterns of activity, but ultimately concluded that the faster patterns more closely resemble the essence of sleep.
If we compare the slow-moving waves traditionally used to define sleep to thousands of people waving at a baseball stadium, these fast-moving patterns are the conversations taking place among just a few people who decide to join in the wave. The conversations are essential to the overall larger wave and are more directly related to the atmosphere in the stadium; the waves are a secondary result of that.
Observe the flicker
As they studied the hyperlocal patterns of activity further, the researchers began to notice another surprising phenomenon.
While looking at models that predict sleep and wakefulness, the researchers noticed what at first seemed like errors: for a split second, the model would detect that one area of the brain was awake, while the rest of the brain remained asleep. They observed the same thing during wakefulness: for a split second, one area would fall asleep while the rest of the brain remained awake. The researchers call these instances “flickering.”
“When we look at the individual times when these neurons fire, [the neurons] “You’re going into a different state,” Schneider said, “and in some cases, these flickers can be localized to certain areas of the brain, and they can be even smaller than that.”
This prompted researchers to investigate what flicker means about sleep function and how it affects behavior during sleep and wakefulness.
“There’s a natural hypothesis there: Let’s say a small part of your brain falls asleep while you’re awake. Does that mean your behavior suddenly looks like you’re asleep? We’re starting to find that that’s often the case,” Schneider said.
The researchers observed the behavior of mice and noticed that when parts of the brain flickered to sleep while the rest of the brain was awake, the mice would pause for a moment, as if they were drowsy. The flickering during sleep (parts of the brain “waking up”) was reflected by the animals twitching while they slept.
Flicker is particularly surprising because it doesn’t follow the well-established rules that dictate the strict cycle in which the brain moves from wakefulness to non-REM sleep to REM sleep.
“We see blinking from wakefulness to REM sleep, blinking from REM sleep to non-REM sleep, and every possible combination that you could imagine, and they break the rules that you’d expect based on 100 years of literature,” Hengen said.
“We think these studies reveal a dissociation between the macro states of sleep and wakefulness at the whole-animal level and the basic state units in the brain: fast patterns and local patterns.”
Impact
A better understanding of the patterns occurring at high frequencies and flickering between wakefulness and sleep will allow us to better study neurodevelopmental and neurodegenerative diseases associated with sleep dysregulation.
Both Hausler and Hengen’s lab groups are interested in understanding this connection further, and Hausler is interested in studying these phenomena further in brain organoid models, which are pieces of brain tissue grown on the bench.
“This could give us a very sharp scalpel to tackle problems in disease and disorder,” Hengen says. “The more we fundamentally understand what sleep and wakefulness are, the better we can address the right clinical and disease-related questions.”
At a fundamental level, this research helps to improve our understanding of the complexity of the brain’s many layers as an organ that determines behavior, emotions, and more.
About this Sleep and Neuroscience Research News
author: Emily Cerf
sauce: University of California, Santa Cruz
contact: Emily Cerf – University of California, Santa Cruz
image: Image courtesy of Neuroscience News
Original Research: The access is closed.
“Non-oscillatory embedding of brain states at millisecond scale provides insight into behavior” David Hausler et al. Nature Neuroscience
Abstract
Non-oscillatory embedding of brain states at millisecond scale provides insight into behavior
The most robust and reliable signatures of brain states are enriched in rhythms between 0.1 Hz and 20 Hz. Here, we explore the possibility that the fundamental unit of brain states may be on the millisecond or micrometer scale.
By analysing high-resolution neural activity recorded over 24 h in 10 mouse brain regions, we show that brain states are reliably discernible (embedded) in fast, non-oscillatory activity.
Sleep and wakefulness states can be classified into 10 stages.0 Up to 1011,000 ms of neural activity sampled from 100 µm of brain tissue. In contrast to standard rhythms, this embedding sustains over 1,000 Hz.
This high-frequency embedding was robust to substates, sharp-wave ripples, and cortical on/off states: individual regions intermittently switched states independently of the rest of the brain, and these brief state discontinuities coincided with brief behavioral discontinuities.
Our findings suggest that the fundamental units of states in the brain coincide with the spatial and temporal scales of neuronal computations.
