# Is there a better way to describe brain activity than EEG “brain waves”

I've been reading about EEG brain waves, which are specific waveforms that are observed on the EEG output, and are usually scored by humans. This concept has been around for quite some time.

Is there anything "newer" or "better" than EEG brainwaves discovered in the recent years?

Given the rise in commercially available EEG sensors, has any company succeeded at putting forth a way to analyze or quantify the output of these sensors to give some useful information to researchers? I'm thinking of brainwave-processing algorithms. For example in Actigraphy, the study of human motion, there are algorithms like "if 19 out of 20 minutes of activity are scored as sleep, then the sleep onset is known to happen at the start of the 20 minute window". Is there something similar for brainwave or derived metrics?

-
Based on the discussion in my answer, the most direct way to clarify is to ask the OP. What do you mean when you say brain activity? Subthreshold activity, spikes? Both? Something else? –  mac389 Oct 9 '12 at 13:51
I should clarify that brainwaves are usually described in terms of their function (ex: if I see beta brain waves on EEG, I can expect that the brain is awake and is busy thinking). I'm interested in brain metrics that allow an observer to reason out something about the brain. –  Alex Stone Oct 9 '12 at 15:37

## 3 Answers

There is no better way to describe brain activity than brain waves! :)

There are newer ways to analyze and think about brain waves, though. Usually you will find these under literature on neuronal oscillations.

Good aspects of thinking about brain activity using brain waves:

• Brain waves are directly related to neural activity. They are an electric or magnetic measure of current passing through neurons. It doesn't get closer than this (unless you use an invasive technique). In comparison, fMRI measures changes in blood flow over one or two seconds after neurons have fired.
• Brain waves are measured essentially at the speed of light, i.e. with no perceptible delay between the onset of neural activity and the moment that the electrode or MEG sensor picks them up. This allows for measuring brain activity as it unfolds, as the brain processes information on the order of milliseconds.

Bad aspects include:

• Seriously compromised spatial resolution. With EEG, the electrical signal gets smeared by passing through the skull. With MEG the measure is more direct, but still, all sensors pick up activity in any part of the brain (at least in EEG; I'm not so sure about MEG), so it is very difficult to say with certainty where precisely something is localized. Methods do exist, but (in my experience) it doesn't come close to MRI precision. It also gets very complicated very quickly.
• Activity is picked up only from the neocortex. You can't say much about deep brain activity.
• Brain waves are utterly unintuitive. It is very hard to see how a wavy line on a screen can tell us something about the most interesting thing in the world: ourselves. It is supposed to be related to our consciousness, to our thoughts and feelings, to our personality, state of mind, actions, intentions, everything - yet it's just a wavy line. This is where some hard training in neuroscience comes in.

In the past, brain waves were identified by eyeballing. Nowadays we can do much better. We can imagine that each of these lines is actually composed of many different types of neural activity, happening at different speeds. Neurons firing together quickly will lead to small, brief amplitude changes, whereas neurons firing together at a slower pace will lead to slow waves. The more neurons are firing together, the higher the amplitude of the wave. The faster they are firing together, the lower the wavelength.

The fast and slow wavelengths will be summed up into the single wavy line on the screen, but we can decompose it, e.g. by using a Fourier transform or a wavelet transform. By decomposing, we gain insight into the fact that neurons fire together at different rates when they do different things. For example, if you ignore a visual stimulus, then your visual brain begins to fire at about 10 times per second. This is called the alpha rhythm. If you are actively observing a stimulus, the activity will change to 40-70 times per second. This is the gamma rhythm. Thus, you can say something about what the brain is doing when you see the pattern of neural activity based on a wavy line.

Also, you might notice that two distant parts of the brain have neurons firing together... but that they are in phase with each other. This might mean that they are both processing different aspects of the same information, so it tells us something about functional neuronal connectivity even though we don't see the underlying white matter firing.

In any case, brain waves should not really be seen as better or worse than other measures of neural activity. They simply add their bit of information, that is part the larger puzzle of how the brain works. See here for a really nice comparison of M/EEG with fMRI, and a discussion of all the big issues behind noninvasive electrophysiological measures.

-
Thank you for an excellent answer! I will look at MEG in more detail, as the term is entirely new to me. –  Alex Stone Oct 9 '12 at 15:42
@AlexStone, you're quite welcome :) About MEG: electricity and magnetism are two aspects of the same phenomenon, electromagnetic force. Whenever electrical current flows down something (such as a dendrite), a magnetic field is generated around it. You can then measure either the electric or the magnetic activity, depending on the equipment you have. The output you get is very similar. (disclaimer: I'm not a technical person, this is just my understanding of how this works) –  Ana Oct 9 '12 at 18:55
Another stupendous answer from you! Thanks for lending us your expertise :) –  Chuck Sherrington Oct 9 '12 at 23:05

Yes and no. Source estimation has been utilized in electrical engineering for decades, but is becoming more and more prevalent in the EEG realm, especially in light of efforts to register EEG readings with concurrent fMRI studies.

Basically, given a set of EEG (or even MEG, magnetoencephalographic) measurements, can we "invert" them to find the individual current sources that would generate such electrical activity. Forward models in which a set of sources are assumed a priori can also be used.

Scholarpedia has an extensive article about these methods. Briefly,

• Parametric dipole modeling
• Spatial scanning and beamforming, one of the more popular being MUSIC
• Sourcespace methods, one of the more popular being sLORETA

are used, most of which involve some form of matrix transformation of the data into a lower dimensional space and mapping on to a set of point sources.

Current research (e.g., Antelis and Minguez, 2012) uses an approach which brings together the results of multiple dynamic models to improve the estimates.

References:

Antelis, J.M., Minguez, J.(2012). DYNAMO: Concurrent dynamic multi-model source localization method for EEG and/or MEG. Journal of Neuroscience Methods, Available online 26 September 2012 [DOI]

-
I will try to flesh a bit more of this out later, but it should give you a head start. –  Chuck Sherrington Oct 9 '12 at 5:17
You might also mention the analysis of oscillatory brain activity. Concerning possible (commercial) applications, one could mention, for instance, Brain Computer Interfaces (BCI). –  H.Muster Oct 9 '12 at 8:43
@H.Muster I was leaving some of that for you! Yes, I suppose that one of the two of us should cover ERS/ERD. I had thought of that as being more "traditional", but I think you are right. –  Chuck Sherrington Oct 9 '12 at 9:09

I'm surprised that no one has mentioned spiking activity. The spatial and temporal resolutions are phenomenal.

The technology to record action potentials simultaneously from many neurons over many cortical areas is growing. Much of theoretical neuroscience deals with how those patterns of spiking convey information.

As with the other answers, I will add to this sketch over the coming days. I wanted to make the point that there are other sources of information about brain activity than EEGs.

-
It's very difficult to discern spiking information within the ensembles using EEG, though. Even if you're talking about using extracellular microelectrodes, the "brain waves" are more the local field potentials and not spikes. That said, I do like the answer in terms of a different interpretation of the main question in the title. –  Chuck Sherrington Oct 9 '12 at 12:53
People do record brain activity using extracellular electrodes. The method for extracting action potentials is very different from getting local field potentials, though. If you record at a high rate, like 40 kHz, and band-pass filter, say between 300 Hz and 10 kHz, then you will see action potentials. –  mac389 Oct 9 '12 at 13:24
Yes, I know, I did extracellular recording for a number of years, but I think it's just a semantic difference, I personally would not refer to a spike to be a "brainwave". I would reserve that for the aggregate activity. –  Chuck Sherrington Oct 9 '12 at 13:29
I totally agree that a spike isn't a brain wave. But, a spike train does approximate a neuron's rate function and so many spike trains from a region can be combined to make a "population rate function". That seems to be something like a brainwave. –  mac389 Oct 9 '12 at 13:32
On people, it's usually done on neurosurgical patients who have electrodes implanted for intractable epilepsy, Parkinson's, psych disorders. They have in animals and do, at a few med centers, flexible electrodes that record a signal. Wireless can't transmit at high enough frequencies to send spike data in real time. The electrodes also have low enough impedance to pick up only a few neurons. (In part because MDs know more about what to do with EEGs and spikes are used for intraoperative localization.) –  mac389 Oct 9 '12 at 15:54