MOST INVENTIONS start their lives as expensive technology in the hands of a few, and trickle down to the masses. But not all those consumers end up using the technology in the way the designer intended. And that can lead to lucrative new opportunities for the manufacturer.
A well-known example is the Microsoft Kinect, which has been hailed as a breakthrough in enabling third party development of interactive applications. Since its original release as a gaming peripheral in 2010, the Kinect has been appropriated by researchers and artists for a variety of purposes.
But sometimes the alternative uses that a device is put to greatly overshadows the original intended purpose. This is the case for an Australian start-up that specialises in interfacing the human brain with a computer or other electronics.
Emotiv’s electroencephalography (EEG) headset was originally envisaged for use as a game controller. However, because the device enables unprecedented access to the functioning of the brain, it has opened up a wave of possibilities for the medical research community, and helped to boost the quality of life of many disabled users.
A well-thought build
Emotiv was founded in Sydney in 2003 by technology entrepreneurs Tan Le, Nam Do and Professor Allen Snyder from the University of Sydney. The firm was assisted by a Commercial Ready Grant from the Federal government, as well as the R&D Tax Offsets scheme.
The original goal of the company was to pioneer more realistic and human computer interfaces using mental cues and unconscious responses from the user. To do this, the company envisioned a low-cost, consumer-friendly, wireless, multi-channel EEG headset.
Emotiv targeted its initial product towards gaming applications, because that promised volume sales and high earning potential, but also because gaming provides a good benchmark for the system’s user-friendliness and robustness during everyday use.
According to Dr Geoff Mackellar, CEO of Emotiv Research and Chief Technology Officer for Emotiv Lifesciences, the headset was developed in-house using an Altium EDA package for layout and simulation.
For brain signal acquisition, Emotiv engineers used a high-quality front-end amplifier system, multiplexed to a 16-bit analogue-to-digital converter (ADC) that passes samples at 2048Hz/channel to a Microchip dSPIC processor.
The data is then processed, filtered and down-sampled to 128Hz/channel and is wirelessly transmitted to a custom USB receiver using an ultra low power Nordic 2.4GHz RF chipset.
A key goal for the headset was to achieve the lowest possible power consumption at the best available price with a good signal-to-noise ratio (SNR) yielding high signal fidelity from micro- to milli-volt levels across a frequency range spanning 0.2 to 45Hz.
The company tested over 50 input amplifier and ADC combinations, and benchmarked the resulting signals against a $30,000 medical EEG system. All wired connections in the system were hard soldered for durability and to eliminate to cost of connectors.
“Where necessary, we used more expensive components but we targeted a low bill of materials (BOM) from the start and worked to hit our target,” Dr Mackellar explained. “We spent a lot of time and money in the design phase to ensure the manufactured cost would be as low as possible.”
To ensure signal fidelity, Emotiv employed a variety of noise mitigation techniques. The primary source of noise in “biosignals” is the common-mode pickup from the 50 or 60-Hz mains together with significant harmonics. To counter this, the system uses a common-mode feedback system where a reference electrode signal is inverted and passed back to the body through a second reference electrode, effectively “floating” the detection circuit on top of the common-mode signal.
This technique, common with high-end medical EEG and electrocardiogram (ECG) systems, provides around 55dB of common mode rejection at 50Hz. Data is collected from each sensor at an elevated frequency, and each channel is then sampled at 2048Hz to capture all of the mains harmonics within the Nyquist range.
Two additional digital filters are then used to eliminate mains-related frequencies and other interference before down-sampling to 128Hz per channel. A standard low-pass filter with a cut-off around 85Hz removes all of the harmonics from the signal without interfering with the brain signal, and a dual-notch digital filter at 50 and 60Hz attenuates residual mains fundamental frequencies below 85dB. As a result, the system maintains a Least Significant Bit sensitivity of 0.5uV enabling it recover brain inputs effectively.
In 2009, after successfully developing the hardware, Emotiv released the developer EPOC headset with 14 saline sensors, wireless communications running a proprietary protocol, and 12 hours of use via a 600-mAh lithium battery.
The headset is capable of detecting four mental states, 13 conscious thoughts and facial expressions (through electromyography, the recording of electrical activity of muscle tissue). In addition, the device uses two gyroscopes to sense head movements.
The headset is complemented by the company’s detection suite software, which is able to read and interpret thoughts and intents, without requiring access to the raw EEG data. The software was intended for people interested in making applications for the device, and resulted in 20 applications being developed.
A community grows
“As soon as we released the developer model, we suddenly got a lot interest from researchers all over the world asking us to make an EEG-enabled product,” said Dr Mackellar. “So we released [the researcher edition] a month or two after the initial developer sales started.”
Today, Emotiv sells applications (‘apps’) for brain-controlled photo viewing, brain mapping, games and brain-controlled keyboards, plus developer tools. And a visit to the company’s support forums provides a glimpse of the buzz of activity surrounding the headset. Carers, researchers and developers post on the board seeking assistance in using the EPOC headset for various purposes.
One thread discusses the use of the system to train children with Down Syndrome, another, the control of a robotic arm, and yet another talks about allowing quadriplegics to communicate and interact by thought.
In addition to more than 20 published papers centred around the device, Dr Mackellar says thousands of research projects are ongoing around the world using the EPOC system.
One notable project is a brain-controlled wheelchair, led by Steve Castelloti, CTO of Puzzlebox, which has implications for the mobility of people with spinal cord injuries. Castelloti’s work is focused on Puzzlebox Brainstorms, an open source software suite aimed at introducing students to basic neuroscience and brain-computer interfaces (BCI).
In 2010, Puzzlebox demonstrated the first version of the system. The user was fitted with the EPOC headset, which transmitted to a module within the Puzzlebox Brainstorms BCI software housed in a laptop computer. An Arduino – the popular open-source single-board microcontroller – was used for setting direction control, while a custom interface box simulated the wheelchair joystick controls.
While the original system allowed the user to move the wheelchair by thinking about the direction they wanted to move, the concentration required to keep the wheelchair in motion, and the distinct thoughts needed to execute simple commands like left and right, created a lot of mental strain just to be able to move around.
Castelloti and his team realised that quadriplegics and paraplegics placed more importance on increasing personal independence and restoring control of bodily functions, and so redesigned the solution into a hybrid wheelchair and “telepresense” robot.
“Our updated design incorporates the latest in computer vision and robotics engineering to provide automated pathfinding and interaction with simple objects using a custom arm,” Castelloti told Electronics News. “Instead of manual navigation, the user is presented with a floorplan or map of their current location. They can make a single selection using a P300-based [BCI] menu and the robot will figure out for them how to best travel there.”
Upon reaching the desired location, another augmented reality overlay is produced, so the user can select various objects, again using the P300-based BCI, which the robot can then retrieve. The robot has freed the user from navigating using constant thought, instead allowing them to use the BCI to make a series of simple selections in order to instruct the robot to carry out various higher-level tasks.
The next brainwave
The popularity of the EPOC can be attributed to Emotiv’s openness to and encouragement of third-party development and research, and the possibility of developing applications of any level of complexity, from simple thought-to-keystroke mapping to full analysis of raw EEG data.
Emotiv claims that the accessibility of the hardware is one of the key reasons for the growth of the development community. For medical-grade EEG studies, a single-channel system can cost between $3000 and $5000, and multi-sensor headsets can cost tens of thousands of dollars.
At these prices the EPOC headsets are less expensive then traditional EEG systems allowing researchers to buy multiple units and execute studies with several people in the same room being measured in parallel to yield faster results.
Having seen the research and development community around the EPOC flourish, Emotiv is now taking the initiative a step further, aiming to leverage the large installed base of headset users to develop a normative database of EEG data.
To encourage users to submit their EEG data anonymously, Emotiv is setting up a wellness website to track users’ mental performance during certain tasks, and provide feedback, so users can optimise their performance.
“Even on the most pessimistic assessments, we have thousands of users with maybe 20 or 50 sessions each. So we would have, in a matter of months, more EEG data on record than any other EEG system in history,” said Dr Mackellar.
The large amounts of anonymous EEG data will become a valuable resource, claims Dr Mackellar. Researchers will be able to quickly sift through the data for example, to look for biomarkers for conditions like anxiety, depression or autism.
Users will be able to participate in research programs, either on a voluntary or paid basis. They can also opt to allow researchers access to their historical data, which will help in searching for precursors to conditions.
“Our research community … will be able to pre-package neuroscience experiments. That could be anything from a stimulus response, or using questionnaires to record EEG data, so they could be used for basic research, or to challenge subjects,” said Dr Mackellar.
“We might demographically sub-divide the groups, so the system will be like a large group of captive researchers, all of whom have their own measurement equipment, and the researchers will be able to … get results within a matter of days.”
The product is already in development, and Emotiv has started pilot trials, capturing the initial part of the brain bank with data from a group of high-achieving individuals.
A tough road to recognition
Emotiv is careful to point out that the EPOC system does not have any approvals for medical use, be it with the US Food and Drug Administration (FDA), or the Australia Therapeutic Goods Administration (TGA), and all therefore all experiments are investigational.
As such, while the headset has potential for use as a remote telehealth system (for example, providing EEG diagnosis capabilities to communities in remote areas of Africa), the company is not marketing it for such applications.
This lack of approval means some researchers using the EPOC system are encountering significant roadblocks in getting their work recognised. One such example is Dr Tim Gureyev, research scientist at CSIRO, and his research collaborators at the Monash Medical Centre.
Dr Gureyev’s engagement with EEG started with his interest in bio-feedback. This is a phenomenon where if a certain parameter of the body is shown to the subject, they can quickly learn to control it.
According to Dr Gureyev, bio-feedback is applicable to things like skin conductivity and heart rate, but there are techniques for bio-feedback relating to brainwaves. “For example, a person can be trained to increase the ‘theta wave’ activity in the left-half of the brain, which may be associated with an improved mood, and [this] can be used [to combat] depression,” he said.
Many of Dr Gureyev’s projects involve extending the applications of EEG. For example, he participated in a recent paper published by researchers from the Monash Medical Centre where a conventional medical-grade EEG headset was used to diagnose stroke.
“There have been some very interesting advances in the treatment of stroke victims,” explained Dr Gureyev. “In some instances, if the drugs are administered early enough, in the three hour window after the episode, in certain types of stroke, it can almost completely reverse the effects.”
The drug, a protein called tissue plasminogen activator (tPA), breaks down blood clots, and is very effective in treating embolic or thrombotic stroke, but is harmful if administered in the event of a haemorrhagic stroke. Thus, a quick diagnosis of the type of stroke which has occurred is vital.
While MRI and CT scans are usually used to diagnose stroke, they are complicated, require the use of large machines, and can be time-consuming, a factor which limits their use when urgent diagnoses are required. The researchers compared the results of using a low-resolution EEG image of brain activity against CT and MRI data, and found the EEG correlated well with the information provided by CT, demonstrating the feasibility of using EEG to quickly diagnose strokes and monitor treatment.
Another of Gureyev’s projects involves developing software for three-dimensional localisation of electrical activity in the brain, from the data collected by EPOC on the surface of the skull. Sometimes called EEG thermography or source localisation, this technique is an extension of EEG imaging.
By finding out where the brain waves are coming from, and isolating particular bands of EEG spectrum, researchers can more specifically associate particular problems or conditions with processes going on in the brain, beyond the raw EEG data.
Gureyev and his colleagues are currently involved in validating the data from EPOC for use in a medical diagnostic context. This is a vital step if researchers working with the headset are to publish any medically-relevant results, or apply for grants from medical funding bodies.
The researchers have already applied for National Health and Medical Research Council (NHMRC) grants in Australia, but their application was rejected because the council did not consider the data from the EPOC headset to be comparable with that of medical-grade EEG devices.
“We need to prove that the data that can be acquired with EPOC is relevant to the target applications, so it can be shown that similar patterns can be seen in the EPOC data as can be seen in the conventional medical EEG devices under similar conditions,” said Gureyev.
According to Emotiv, the U.S. medical research community seems much more receptive to the use of the EPOC, and it is in conversation with the FDA to get an investigational device exemption for research work in children’s hospitals.
The company is also involved in some comparative and benchmarking studies, which it hopes will lead to some published papers that will prove the capabilities of the headset to the Australian authorities.
But it seems inevitable that if the EPOC is to be recognised as a serious tool of research, it will need to go through the lengthy and costly process of being approved for medical use. And as Emotiv continues building its collection of EEG data, it is almost certain to encounter privacy and data security issues, especially because of the personal nature of the data.
It remains to be seen if Emotiv and its research community can successfully address these challenges, but one fact remains – through its creation of an effective and accessible EEG headset the team at Emotiv has already given thousands of people unprecedented access to the workings of the brain that may help the disabled live more productive lives, improve the accessibility of EEG to remote, poor areas and improve outcomes for stroke victims. That’s not bad for a product that was originally designed for gamers.