Viewing Public EEG Data With EDFbrowser

Discovering the effectiveness of free EEG tools while learning more about BCI

An EEG headset being set up

On October 2nd, I started my quest to learn to read EEG data.

But to further understand the concept, I needed to get my hands on some raw EEG data, and play with it myself.

It should be easy right? With all the free and open source resources available, I just need to download some software and datasets, I thought, as I began this project.

I have never been more wrong. Instead of the easy plan I had in my head, I went down a rabbit hole of wrong file types and windows defender messages (you know, this one?).

I spent days looking for a way I could view data without having to pay for an EEG or the software, because purchasing either one would have really hurt my wallet.

Finally, on my fifth day of countless hours searching, I found that I could download another piece of software that would not only upset Windows Defender, but also read EDF/BDF files (typically used for MATLAB), titled EDFbrowser.

But First, a little background on BCI and EEG

In recent years, EEG headsets have become more popular and budget friendly, with the Neurosky Mindwave starting at $70 and the Muse Headset at $250 (at time of writing). However, more precise headsets with more electrodes cost significantly more.

All EEG headsets use the very important international 10–20 model.

This model provides a guide for setting up electrodes and fixed headsets (EEGs with connected, fixed electrodes). Each electrode sits over a specific part of the scalp, because that electrode typically correlates with an area of the brain.

If I were to explain every electrode and what area of the brain they correlated to, this article would be extremely long. A great resource for this is linked here.

However, the 10–20 system has its flaws. It is extremely difficult to replicate exactly where electrodes are on different people’s heads. Additionally, as BCI tech has advanced over the years, the 10–20 system has not allowed enough electrode positions for more channels.

Why EDFbrowser?

Other than that, there’s not much to it. It’s by far the easiest application I have come across to use, but at the same time it’s rather difficult. There is very limited documentation and nearly no guides on how to use the software, so it requires an inexperienced user like me to push random buttons until they figure out what happens. At least I am good at that right?

That being said, it didn’t take very long to figure out.

But for some reason, the wrong signals kept getting deleted, and most datasets’ files didn’t work very well.

Errors between signal colors and organization

Colors didn’t match the signals (an error from there being far too many values), and the work of manually fixing the issue took an hour per file — far too much time, considering I would want to test many datasets.

Physionet.org’s Arithmetic Task Dataset

During the study, participants would wear an EEG headset prior to doing anything as a baseline. Then, participants would wear the headset while they performed a simple subtraction problem.

Finally! When adding the range of signals to EDFbrowser, the signals matched and arranged themselves perfectly. It was about time to get some success (this was around 1 month after starting the project, and I was getting frustrated to say the very least).

Reading the data

It’s also important to know that natural “squiggles” can simply just be background activity in the brain (delta and theta signals, more information here).

So what can be learned from this study?

Starting with the subject 1, big differences can be seen when the subject does their tasks.

Subject 1’s spike, while doing tasks (see below)

Big spikes occur typically from the O2 region, a node over the Secondary Visual Cortex (responsible for helping with visual information). Could this be the participant reading the problem and interpreting it? While there is no way for someone as experienced as myself to know, it’s an appropriate guess.

More of subject 1’s spikes while doing tasks (see below)

Another big spike happened here, but instead of the O2 signal being the star of the show, P4 and Pz take very similar spikes. The P4 signal is over the Angular Gyrus, which can be responsible for number processing and reasoning. Just like the last example, it’s likely caused by doing a problem. The Pz signal, however, rests over the Parietal Lobe, which is most likely (not entirely confirmed), responsible for manipulating information in the working memory.

Other participants didn’t see much change at all while doing their tasks.

By looking at these two images of data, both from the same participant, is it simple to tell which one is from the baseline? Definitely not.

In fact, some data looked like the brain had more activity while doing nothing at all.

Subject 15’s Baseline
Subject 15’s Arithmetic Task Data

Does the baseline lack focus? Is the subject panicking from having the headset on their head? There is no way I could tell. This is a prime example of how BCI data can go how it is least expected to.

It’s important to realize that I am no EEG specialist, because somebody with past experience, who was present at the time of the recording could have learned a lot more than I did. Knowing when someone starts a new problem, or spends a while on one of their tasks is very important. Obviously, a video of the subjects can’t be added for anonymity reasons.

So what can be taken away?

  • Sometimes EEG data is clear and easy to read/interpret
  • Sometimes EEG data is extremely confusing and difficult to interpret
  • Most BCI tech is behind a paywall (understandably so)
  • Resources for learning about BCI that are clear can be hard to find, but they are buried within the internet if someone looks deep enough.

Other:

  • Windows Defender is no longer happy with me
  • I have no idea if my computer now has viruses
  • BCI data is really, really amazing.

Is BCI the key to our brains? It’s not clear yet, much like my interpretations of data. Either way, BCI and EEG are going to become a much larger part of our lives, and we better strap in for an exciting ride.