Using innovative sensor technology, IMEC researchers have demonstrated a headset that captures brain signals.
Imagine a comfortable and lightweight headset that you put on and that immediately starts to capture your brain signals. No wires and additional electronics, everything is in the headset: a complete, flexible EEG (electro-encephalogram) system.
The recorded brainwaves are processed in the headset, depending on the application. And the results are sent wirelessly to a base station, for example a doctor’s computer, in the case of a clinical application, or a game console for an entertainment application. The headset is easy to set up; there are no lengthy customization procedures. It adapts to the form of your head for the best possible scanning results. Available at your local pharmacy, the EEG headset works for weeks or even months at a time without battery recharges.
Such a headset could eventually replace today’s EEG systems, including the scanners at the hospitals, as well as those used for ambulatory EEG scanning. The latter, used to monitor epileptic patients in their home environment, for example, are still quite bulky. Next to a headset, they require wired connections to a monitoring box the size of a laptop. For patients, it’s the hospital imported in their homes, still far from a comfortable monitor that they can install and forget.
But this future headset would also allow for the design of new applications. Think of drowsiness monitoring in cars and trucks, inexpensive high-quality sleep monitoring at home, biofeedback applications for cognitive therapies (treating anxiety disorders, addictions), or emotion monitoring applications in the workplace. There’s also the whole application field of the brain-computer interface—steering applications through interaction with the mind rather than of the body.
How far are we from having a comfortable wireless EEG headset? There are some products available today, mainly for gaming applications.1 They all implement part of the requirements of our visionary headset, but not all at the same time. What’s key to these systems is that they introduce new possibilities and applications, and stimulate further research into low-power electronics and application integration.
One of the most important issues to solve remains the autonomy, or use time, especially when a wireless radio is employed. Currently, the batteries of most systems run out after maximum 12 hours. Although this may be enough for some applications, carefree ambulatory monitoring requires an autonomy measured in weeks or even months, not hours. The current applications are made with off-the-shelf components, so there’s a power consumption gap of one to two orders of magnitude between the low-power electronics that are currently available commercially and what’s really needed for a wireless EEG headset or comparable body sensors.
Another issue is intelligence. Most existing systems scan raw data and send these to a base station or computer for filtering and analysis, instead of doing local processing on the sensor. Note that the quality of biomedical signals generally depends largely on the ability to filter out artifacts, or glitches in the signal, which could be caused by body movements of the wearer. These artifacts are best detected and filtered out directly on the sensor. But that requires extra measurements and processing, adding considerably to the low-power challenge.
And then there is the challenge of miniaturization. The end product must consist of small, lightweight sensors that people can carry comfortably, even unnoticed. Think about the bus driver who would be required to continuously wear a drowsiness monitor headset for eight hours. Any pressure, weight, or discomfort will soon cause the monitor to be discarded. A last challenge will be to make these electronics available for everyone, mass-manufacturing them at a low cost, possibly in a one-chip solution.
At IMEC and Holst Centre,2 scientists are working on the base technology that will underlie future body sensors, such as the EEG headset presented here. In fact, they’ve even built demonstration units (see Figure 1). With every generation, these come closer to the vision of the comfortable, autonomous, user-centered sensors.
Figure 1. The demonstration headset designed by researchers at IMEC is pictured here.
The latest demo unit from IMEC is a headset-only EEG system where all the electronics, including biopotential ASIC, radio, controller chip, and power circuit, are housed in a small (25 by 35 by 5 mm) wireless EEG package. The autonomy currently stands at 1.5 to 4 days, depending on the use. One of the breakthroughs included with this version is the use of dry electrodes, which makes the headset suited for unassisted home use.
Among the many biological parameters that can be measured by sensors, EEG signals stand out because they are particularly challenging to measure. They have a very small amplitude (typically between 1 and 20 µV) and a wide frequency range (from 0.1 to 100 Hz). To record such small amplitudes, the sensor needs a low noise floor. The demonstration headset is designed with an 8-channel ultralow-power analog readout ASIC that was designed specifically to capture EEG signals. This chip consumes only 200 µW and features a high common mode rejection ratio (CMRR) of 120 dB and low noise (input referred noise of 55 nV/√Hz). These performances are achieved at an input impedance of 1G Ω.
Ease-of-use and unassisted setup are key requirements if we want headsets like these to be used for long-term monitoring. Standard EEG equipment uses a gel to get good contact between the electrodes and the wearer’s head. But this can be messy and requires assistance. The dry electrodes make good skin contact without any gel. However, using such dry electrodes causes a high electrode offset and a higher impedance. Hence, these additional challenges were overcome in the amplifier design. Optionally, the electrodes can be injected with gel, when low contact impedance is needed.
To arrive at a maximum autonomy, components were designed and selected that have low quiescent and leakage currents. Parts of the system are shut down when they aren’t needed, such as the impedance measurement. And data can be moved from the ASIC to the microcontroller (MCU) memory while the MCU is in a low power mode. The MCU is activated from the low-power mode through interrupts. Last, the radio is duty cycled.
The signal-to-noise ratio of this monitor is 25 dB on real EEG signals. The entire system uses 3.3 mW for continuous recording and wireless transmission of one channel sampled at 256 Hz, and 9.2 mW for eight channels sampled at 1024 Hz. This gives the desired 1.5 to 4 days of autonomy on a small 100 mAh Li-ion battery, depending on the mode of operation.
A number of applications were constructed to demonstrate and test the IMEC ultra-low-power biomedical sensors, such as the EEG headset. Such demonstrations illustrate what such sensors could mean in future patient care and other health and wellness applications. One application that included an EEG sensor was a comfortable sleep staging monitor, a device that can check a patient’s stages of sleep at home, instead of in a hospital’s sleep laboratory.3 On the headset are three sensor nodes measuring two EEG channels, two EOG channels (electro-oculogram), and one EMG channel (electromyogram). This system was validated in the sleep laboratory at the University Hospital Center in Charleroi (Belgium), against a commercially available reference system, proving that wireless headsets could replace current monitoring systems to monitor sleep stages.
Another application of the EEG sensor was an installation created by the artist Christoph De Boeck, appropriately titled “Steel Sky.” De Boeck has attached 80 steel tiles to a large ceiling. On the back of each tile is a metal pin and dampener. Each metal pin can be driven separately to tick against the tiles, and the sound can be further modified by the dampeners. Visitors wear an EEG headset that detects their brain waves and sends them to a computer that’s connected to the pins and dampeners (see Figure 2). Then, the brain signals drive the pins, resulting in a rising and falling soundscape.
Figure 2. The EEG headset is shown in use in the “Steel Sky” application.
A third application is the mind speller, a headset that interprets brain waves to help the wearer spell words and phrases.4 The wearer sits in front of a monitor that flashes alternate rows and columns of characters. If he recognizes the character that he intends to use, the EEG monitor picks up the characteristic brain waves associated with this recognition. This system demonstrates a portable, easy-to-wear, intelligent textual and verbal communications device enabling people with motoric disabilities (suffering from for example brain paralysis or speech or language disorders) to communicate.
1 Julien Penders, Chris van Hoof, and Bert Gyselinckx, Bio-Medical Application of WBAN: Trends and Examples, in: Bio-Medical CMOS ICs, Springer 2010, pp279-302
2 www.imec.be; www.holstcentre.com