Cutting through the “noise”

November 22nd, 2022

Written by: Margaret Gardner

Most people tend to avoid or ignore static. A static-y signal on TV might mean a storm knocked out your cable, or maybe you use white noise to drown out the sounds of your upstairs neighbors so you can focus. Neuroscientists, too, have often avoided or ignored what they thought was static in recordings of brain activity. But what if this static was actually shaped by something important?

Neurons send information to each other via electrical signals that often combine and travel throughout the brain as brain waves. Just like radio waves, sound waves, or waves at the beach, brain waves can have different properties that describe their shape and how much energy they’re carrying. Frequency is the number of waves that pass  in a given period of time. If you’re at the beach, frequency might describe how many waves hit your toes for every second you’re standing on the shore. Brain waves also have power, which is how quickly something applies energy. The easiest way to think about power is to imagine you need to pick up something heavy; it will take the same amount of energy for you to lift whether you do it fast or slow, but picking it up quickly will take more power. For a wave at the ocean, this would be how rapidly a wave levels your sandcastle¹. Concepts like frequency and power can help sort waves into groups that of waves that have similar shapes, which you can read all about here, and show which of those groups contribute the most to brain activity.

There’s a lot of brain activity that isn’t part of these major brain waves. For instance, if you look at a chart showing how powerful brain waves are at each frequency to see which signals are the strongest, you’ll see what looks like a very steep hill with bumps on it (Figure 1). These bumps are the brain waves, which tend to have specific frequencies in the brain and are relatively easy to distinguish from the “background” signal, which follows the shape of the dotted curve. This curve on the power-frequency spectrum is called the 1/f signal (pronounced “one over F”). Interestingly, this curved shape is seen in many different types of brain signal measurements, from large-scale measures of the whole brain’s activity all the way down to the activity of individual neurons². In fact, 1/f is measured in lots of non-brain data as well, popping up in everything from the shifting of the tectonic plates that make up Earth’s surface to stock market fluctuations to reaction times^2,3. Yet, since 1/f is everywhere and constant across time, scientists figured it was just noise. For a long time, this data was actually called “1/f noise” or “pink noise” and was removed from measurements so that brain waves, aka the “important” stuff, were easier to see². The name “pink noise” may remind you of the term “white noise”, but they’re actually two different kinds of static. In white noise, waves of all frequencies (in sound, pitches) have the same power (volume intensity) (Figure 1, Audio 1). Pink noise, on the other hand, follows the 1/f curve, with low-frequency (low-pitch) waves having more power (volume) than high-frequency (high-pitch) waves (Figure 1, Audio 2). You can both see and hear the difference!

Audio 1. White noise recording.

Audio 2. Pink noise recording.

While there is still a lot we don’t know about 1/f signal and where it comes from, most neuroscientists today seem to agree that it’s not just empty static. When measuring the 1/f component of brain signals, scientists found that the higher-frequency waves are linked to the lower-frequency waves². When the lower-frequency waves have a certain offset, or phase, from each other, the higher-frequency waves get more powerful ². This linkage isn’t there, though, if you just randomly make up fake data points that fluctuate around the 1/f line ²; even though to us they look the same as 1/f brain signal, the “low frequency” fake data won’t be related to the “high frequency” fake data at all, since everything’s random. This shows that even though the 1/f signal in the brain may look like noise, there’s something non-random underlying it². Researchers suggest that this link may be driven by tiny, occasional blips of negative electrical activity that seem to sync up the alignment and increase the power of brain waves². Regardless of what links the low-frequency and high-freqency brain wave signals, though, the fact that a link exists suggests that the 1/f shape is real brain signal and not just from random, useless noise.

If it isn’t just noise, why do so many kinds of data seem to follow this 1/f curve? In each case, from fMRI to heartbeats³, it’s likely that a lot of tiny processes add up to create this familiar shape². This is like how many individual people talking adds up to create a familiar background hum of conversation that sounds similar in any coffee shop. For example, the same scientists who showed that negative blips might sync up brain activity across the 1/f curve note that a similar process seems to happen to seismic waves. Rather than negative electrical activity, these waves are jolted into alignment when tectonic plates bump into each other, showing how different processes in nature could coincidentally be producing similar-looking signals². But even though these negative blips align the brain activity that makes up 1/f signal, they don’t fully explain where that activity is coming from. So, to revisit our coffee shop analogy, what tiny process (or “voice”) is making 1/f signal in the brain?

The current theory is that the 1/f signal shows the balance of all the inhibitory and excitatory signals between neurons in an area⁴. All the electrical signals that make brain waves come from two different types of neurons: excitatory or inhibitory. Excitatory neurons’ signals make the receiving neuron more likely to send a signal of their own, while inhibitory neurons tell the next neuron to pause on sending any signals. The steepness of the 1/f signal (Figure 2) changes to reflect the balance of these excitatory and inhibitory signals, or E:I ratio. As the E:I ratio gets bigger, with more excitation relative to inhibition in the brain, the slope of the 1/f line gets less steep⁴. For example, your brain has more excitatory activity when you’re awake than when you’re asleep or under anesthesia, which scientists have seen to also be reflected in steeper 1/f curves during sleep and anesthesia compared to being awake⁵. Experiments like this, as well as measuring how the slope of 1/f differs in brain areas that have more or less excitatory vs. inhibitory connections, support the idea that the 1/f slope comes from many neurons’ signals adding together and cancelling each other out⁴.

Using the 1/f signal’s curve, scientists can now more easily study situations where the E:I ratio varies. When doing a mental task, for instance, the downward curve of the 1/f signal gets less steep than when the brain is resting, and the curve gets even less steep as the difficulty of the task increases². This could mean that the balance of brain activity shifts more towards excitation when you need to think hard about something. Other studies have found that the 1/f curve’s slope can also be changed by age, diseases like ADHD or schizophrenia, drugs, or the type of measurement tool used^4,5.

Knowing that the 1/f slope is associated with E:I ratio can help scientists interpret their findings and connect how tiny signals from individual neurons are contributing to broader phenomena. Building on this, scientists are working to understand exactly how and why E:I ratio takes the 1/f shape, and what useful information it might hold about brain function and disease. In the end, this so-called “noise” may have a lot of interesting things to tell us.

References

1.        Energy vs power – Energy Education. Accessed November 13, 2022. https://energyeducation.ca/encyclopedia/Energy_vs_power

2.        He BJ. Scale-free brain activity: past, present and future. Trends Cogn Sci. 2014;18(9):480. doi:10.1016/J.TICS.2014.04.003

3.        Ward LM, Greenwood PE. 1/f noise. Scholarpedia. 2007;2(12):1537. doi:10.4249/SCHOLARPEDIA.1537

4.        Gao R, Peterson EJ, Voytek B. Inferring synaptic excitation/inhibition balance from field potentials. Neuroimage. 2017;158:70-78. doi:10.1016/J.NEUROIMAGE.2017.06.078

5.        Gerster M, Waterstraat G, Litvak V, et al. Separating Neural Oscillations from Aperiodic 1/f Activity: Challenges and Recommendations. Neuroinformatics. 2022;20(4):991. doi:10.1007/S12021-022-09581-8

Cover photo made in Canva.

Figure 1 and 2 made with Biorender.com.

Audio 1. Sound Effect by BlenderTimer from Pixabay.

Audio 2. Sound Effect from Pixabay.