Lab Talk

emotiv on neuroscan

Is the Emotiv EPOC signal quality good enough for research?

The Emotiv EPOC at less than $2000 opens up enormous possibilities to take research out of the lab to get access to subjects and scale experiments substantially.  But is the signal quality good enough? 

When Emotiv came out with its EEG EPOC headset it was originally positioned as a gaming system – allowing simple brain computer interface applications for video games. It was a toy for gamers. Dr. Nicholas Badcock and his group at Macquarie University in Sydney Australia were among the first to see its potential in other applications.

Nic did his PhD in developmental psychology and then studied language processing at Oxford before returning to Australia. He now studies auditory processing in children with dyslexia. As any researcher with a substantial grant would do, his team purchased the upscale Neuroscan EEG system for their work. The Neuroscan system is tens of thousands of dollars but with good contact quality and a 1kHz sampling rate. It records a high quality signal – it’s the gold standard. It would all have been just peachy except that Nic and his team were faced with some significant challenges when it came to recruiting subjects for their work. Only 5-10% of the population has some form of dyslexia and they are spread out and hard to find. Moreover, getting these kids to come into the lab from all over was no simple task. What if they could take their lab into the schools or homes of subjects instead? Not only would this allow them to reach a much larger number of children but also potentially enable early diagnosis and intervention for kids with dyslexia. This was not possible with the Neuroscan since it takes time to set up and is not easily portable (see Taking Neurotechnology out of the Lab for a view on EEG devices in general).

The Emotiv EPOC on the other hand is easily portable and quick to set up (read more about its function here). Also appealing is the price point. At less than $2,000 you could buy ten of them and easily scale the research to a larger number of subjects. The obvious issue of course was signal quality. Would it be good enough to be useful? Most people would have simply dismissed it as not even a possibility. The first reactions are that contact quality probably isn’t good enough so its probably really noisy and the saline sponges could dry out which would result in the contact quality degrading causing the impedance to increase beyond use and the baseline to drift. As we once heard one person say – ‘Yeah it probably measures something, who knows what.’ But really you don’t know until you try and with the fairly substantial skepticism out there its not simply seeing if you can get a result but directly comparing the result to the gold standard. Until then, everything you measure and report with the EPOC would come with a pallor of doubt. Nic took it head on. When the reasons are compelling and the motivation high…. (Thank you Nic for the pioneering effort!).

emotic_on_neuroscan_kids

The Set-up

Their slick set up involved cutting out holes in the Neuroscan cap to fit the Emotiv EPOC electrodes in positions close to the Neuroscan electrodes, and syncing the two systems to the same stimulus trigger with a bit of software modification.

 

neuroscan_emotiv_channel_locationsnic-set-up-2

 

 

Each position in the brain will produce slightly different activity so you can’t just look to see if the traces match up. Rather they looked at the response to auditory tones. This is an increase in the amplitude of the trace that arises and peaks at about 300 ms after a stimulus is presented (called the P300 or P3). It’s a pretty consistent response and is fairly similar in neighboring regions.

The Results

So how did it match up? Pretty damn well in the 18 subjects they tested it out on. The P3 was not significantly different between the two when they compared the two closest electrode pairs in both systems. Take a look at some of the traces.

nic-results

It would be surprising though if they had found no differences at all. In some cases they report that the peaks, mainly the earlier smaller peaks (P1, N1, P2, N2), were detected about 8 to 15 ms later in the EPOC, which is a likely result of the 8x lower sampling frequency. They also found that some of the earlier peaks (P1, N1, P2, N2) were sometimes statistically smaller in amplitude in the EPOC – this would be on account of higher noise due to poorer contact quality. However, all in all it was plenty good enough to find what they were looking. You may miss some small differences with the EPOC (false negatives) but the differences you do pick up would be real.

The ERP is one measure and it doesn’t test everything of course. There are other aspects that may not bear out as well. For example, if the person moves around a lot it would likely pick up more movement artifacts more stronglt or if you were to do very long experiments (>20 mins) the signal quality would degrade as the contacts dry out. However, with your subject sitting pretty still in a chair and with experiments that are under 20 minutes, research away!

You can watch Dr. Nicholas Badcock talk about his work here and read the original paper here.

3 thoughts on “Is the Emotiv EPOC signal quality good enough for research?

  1. Interesting findings! The number of trials per condition is a question. High numbers to an extent can compensate for a low SNR, which works for P3 but not lower signal earlier deflections. An analysis of how many trials are needed to eliminate differences in P3 amplitude or prominence between systems would be instructive to low-budget investigators. The saline electrodes, not EPOC’s electronics, are the main drawback affecting these earlier deflections. Here is an interesting solution: https://github.com/SmartphoneBrainScanner/smartphonebrainscanner2-hardware/blob/master/easycap/build_instructions.md

  2. Here is a research paper by Jeremy Frey, comparing OpenBCI to a medical grade device for BCI applications. OpenBCI does quite well in the comparison,

    https://hal.inria.fr/hal-01328427/document

    Also note that OpenBCI samples at 250 Hz, when used in 8 channel mode. Higher sample rates are possible using an opto isolated wired USB link, documented on our forum.

    Regards,

Leave a Reply