Time-Frequency Analysis for EEG/MEG Explained! | Neuroscience Methods 101

Time-frequency analysis is a way to analyze signals from electroencephalography (EEG) and magnetoencephalography (MEG). But how does it work?
In a time-frequency analysis a brain waves are analyzed. The brain signals we pick up with EEG reflect group level activity of neurons below the sensor. This activity is often oscillatory. These brain oscillations have different functions depending on their frequency. As such, we would like to see how brain oscillation frequency changes over time and how it changes when we are doing a task. In that case we look at event-related oscillations (ERO)
With time-frequency analysis we can look at evoked activity, which is when a phase reset of the oscillation occurs. In this case event-related oscillation is often similar to the event-related potential (ERP), which is the EEG signal in the time (temporal) domain. Indeed, EROs are expressed in the frequency domain.
We can also look at induced activity, where a change in brain oscillation power is not associated with a phase reset. ERPs will often look flat, because the signals cancel each other out. However, time frequency analysis will still show a clear response.
References:
Adjamian P. (2014). The application of electro- and magneto-encephalography in tinnitus research - methods and interpretations. Frontiers in neurology, 5, 228. doi.org/10.3389/fneur.2014.00228
David, O., Kilner, J. M., & Friston, K. J. (2006). Mechanisms of evoked and induced responses in MEG/EEG. NeuroImage, 31(4), 1580-1591. doi.org/10.1016/j.neuroimage....
Herrmann, C. S., & Knight, R. T. (2001). Mechanisms of human attention: event-related potentials and oscillations. Neuroscience and biobehavioral reviews, 25(6), 465-476. doi.org/10.1016/s0149-7634(01...
Herrmann, C. S., Rach, S., Vosskuhl, J., & Strüber, D. (2014). Time-frequency analysis of event-related potentials: a brief tutorial. Brain topography, 27(4), 438-450. doi.org/10.1007/s10548-013-03...
Magazzini, L., & Singh, K. D. (2018). Spatial attention modulates visual gamma oscillations across the human ventral stream. NeuroImage, 166, 219-229. doi.org/10.1016/j.neuroimage....
Rosenblum, Y., Shiner, T., Bregman, N., Fahoum, F., Giladi, N., Maidan, I., & Mirelman, A. (2022). Event-related oscillations differentiate between cognitive, motor and visual impairments. Journal of neurology, 269(7), 3529-3540. doi.org/10.1007/s00415-021-10...
Wischnewski, M., & Schutter, D. J. L. G. (2019). Electrophysiological correlates of prediction formation in anticipation of reward- and punishment-related feedback signals. Psychophysiology, 56(8), e13379. doi.org/10.1111/psyp.13379
Narrated by: Miles Wischnewski
USE OF STOCK IMAGES AND VIDEOS
Occasionally we make use of stock images and videos (pixabay.com or pexels.com). We use these to make our videos more lively (better than looking at a blank screen). These stock videos should not be taken as an exact scientific reflection of the discussed content. In some cases they are not fully accurate. We try to have a high viewing pleasure, while at the same time create as little confusion as possible. Thus we aim to only use them when it is clear that they are not directly related to the discussed content.
CHANNEL GOALS
The Psyched! channel focuses on providing education about psychology and neuroscience. Here you learn all about human behavior, the mind and the brain. It is our goal to reach everyone, from people who are casually interested to academic researchers. As such we provide simple explanation videos as well as profound lectures. If you like our content, consider to subscribe!
SOCIAL MEDIA LINKS
Instagram: / real.psyched
Twitter: / mileswischnews1
Website: www.mileswischnewski.com

[@dhandapani9870]

Thank you so much. You’re such a blessing

[@BioniChaos]

Great explanation, thank you!

[@hubandad]

This video helps me get clear the difference between the evoked potential and induced potential. Thx a lot man.

[@real.psyched]

Our pleasure! Glad it was helpful!

[@gussy-ps4he]

thanks!..pretty clear!

[@jansirani4429]

Very informative

[@brianwilkinson5752]

Hey, I was doing some research on using the Fourier transform on EEG data and it would be nice if we could decipher the non-sinusoidal waves associated with single firing neurons within the pool of firing neurons in an EEG recording. The problem with using the Fourier transform on single neuron sites is that it solves for the individual sine waves when single neuron waves are non-sinusoidal. I came up with an idea that may make it possible to decipher the neuron non-sinusoidal brainwaves at each location within the brain. My idea is that you would need at least three sensors producing EEG data simultaneously. For each sensor, you could apply the Fourier transform and decipher all the individual sine waves. If each sensor detects all the waves of every single neuron firing an action potential in the brain at some particular time, then theoretically, each of the three + sensors should produce the same individual calculated sine waves in the frequency domain plot at some particular time, just at varying amplitudes of each calculated sine wave for each sensor since the three + sensors would be located at different locations at the scalp. If that does turn out to be the case, then each calculated sine wave associated with its frequency could be located to a specific location within the brain by identifying the amplitude of each sine wave with their corresponding frequencies in each of the three + EEG recordings. Another words, identify one of the calculated sine waves in one EEG sensor recording and then find this same calculated sine wave in the other two + EEG sensor recordings and then determine the location within the brain of the calculated sine wave based on the amplitude differences of the calculated sine wave in each of the three + sensors. This would be done for all the calculated sine waves in the EEG recordings of all the sensors. Next, combine all the calculated sine waves that correspond to the same location within the brain to get the non-sinusoidal shape associated with the neuron(s) at each precise location within the brain. It might be good to know the typical spacing of neurons in the brain, so you don’t add up two sine waves from separate neighboring firing neurons in the brain when you wish to identify the wave associated with each separate firing neuron in the brain. With this idea, it may be possible to determine the individual non-sinusoidal brainwaves associated with single firing neurons by using at least three EEG recording sensors placed at different locations on the scalp. I hope this was clear. Is this a good idea? Has this idea already been attempted? What do you think?

[@real.psyched]

It is a little hard for me to wrao my head around that completely, and I am certainly no expert. I what I do know is that there are methods to specifically look at the non-sinusoidal components of a frequency spectrum (IRASA or FOOOF). But I am not sure if they achieve what you are looking for.

[@brianwilkinson5752]

I did research on that, and it is mostly over my head. Does that mean that we can obtain the non-sinusoidal brain waves of single firing neurons? I am inspired by a vision that perhaps one day, I may be able to use an EEG that will be able to display on a computer screen some image that I am looking at or thinking about and play audio that I am listening to or thinking about. Do you think that is a possibility in the near future?@@real.psyched

[@real.psyched]

Single neuron activity with EEG is generally not possible (neither sinusoidal nor non sinusoidal)

[@response2u]

Thank you sir! I would nice to see a video on lead field in MEG.

[@real.psyched]

Thanks for the tip! More methods 101 videos are coming up! I will put that one on the list! :)

[@vecyl]

I need this!!

[@AranyakBanerjee]

Thaks for such an informative video, but I have doubt while calculating PSD of alpha wave why in the graph the frequecy axis is showing in range of 100-500 Hz

[@real.psyched]

Not sure what you are referring too. As far as I can tell no frequency of 100-500Hz is shown in any plot