It’s not what it looks like: High Frequency Oscillations

Following on from the last post on baseline shifts, this is the second post on a few things we don’t normally consider when talking about standard EEG measurements. I had the idea for this one after a few excellent talks by Liset Menendez de la Prida at ISWP7 earlier this year.

Fast and furious (or is it)

Fast stuff on the EEG is difficult to see for a number of reasons: i) We usually filter raw EEG signal to make it look neater and often exclude high-gamma range signal. ii) The signal we measure on the scalp itself is already attenuated by passing through different tissues, making fast activity appear less sharp and prominent. This is true even for ECOG when compared to direct LFP recordings (which is becoming more relevant now that microelectrodes are being used more and more in patients with epilepsy). iii) Higher frequencies have a lower power – usually fast fluctuations are a lot smaller than bigger shifts on the EEG and seem to pale in comparison, when visually analysing the EEG.

Yet the high frequency components of the EEG seem to be important. When mapping fast activity in patients with focal epilepsy, it seems high frequency oscillations are associated with the seizure onset zone, even in patients where this was outside of a macroscopically identified structural abnormality [1]. These results are similar in their statistical association with seizure onset zones as more clearly pathological spikes.

Now this is interesting, because high frequency oscillations themselves are not always pathological. High gamma range activity (50-80Hz) is commonly associated with exploratory behaviour in neocortex and hippocampus. Faster activity in the hippocampus is known as ripple (80-200Hz)  and associated with memory consolidation. An even faster short burst if high frequency oscillations (250-800Hz) is known as fast ripple.

In a feat of hippocampal electrophysiology recordings using 4-by-4-pronged microelectrodes, Liset Mendendez de la Prida’s lab managed to record different high frequency events and estimate the contributions to the overall event by individual neurons. What emerges is that physiological high frequency events were caused by smaller sets of slightly asynchronous pyramidal cell neurons firing, whilst pathological-like events were so synchronised that spike-sorting was made near impossible with each spike representing a large majority of the neuronal population firing exactly at the same time.

Patch-clamp recordings from individual neurons (top row), local field potential traces (middle row) and frequency time plot (bottom row) of physiological-like and pathological-like high frequency events. Note the association between single neuron firing and LFP trace in the pathological-like events. One can also see that the physiological-like event is less tightly defined in the frequency-range and ‘bleeds’ more into other frequencies than the very stereotyped, single frequency pathological-like event [2].

Some of these differences leave a trace in the raw recording, and someone with sufficient expertise (not me) could talk you through how to recognise the markers of pathological vs physiological high frequency oscillations on the electrophysiological trace of your choice.

What next

Despite them being an exciting field of research, there are lots of barriers to properly using high frequency information in day-to-day EEG recordings. However, technology, particularly of intracranial recordings is advancing, and with the increasing access towards epilepsy surgery, advanced human electrophysiology will become more and more relevant for accurate diagnosis and localisation of surgical targets.

A Mayo clinic group with a custom made depth electrode that contains both (standard) macro-contacts and (experimental) microelectrodes has shown that a significant number of high frequency events is only visible on the microelectrode trace.


On the left you can see the standard depth electrode with little microwires sitting on top. The top (red) lines are microwire recordings and the horizontal bars highlight high frequency events not visible on the macroelectrodes (below). You can also see the DC-shift associated with some epileptic events we’ve talked about in the last blog-post. [3]
Clearly we are getting much more informative data out of these electrodes. When it comes to understanding what goes awry in the circuitry of the epileptic brain, the fidelity of these data has substantial power to change the way we model and understand the neuronal processes underlying these fast and sometimes furious high frequency events.

[1] Jacobs J, et al. (2009)Brain 132: 1022-37. doi:10.1093/brain/awn351
[2] Aivar P, et al. (2014)J Neurosci 34: 2989-3004. doi:10.1523/JNEUROSCI.2826-13.2014
[3] Worrell G, et al. (2008)Brain 131: 928-37. doi:10.1093/brain/awn006