Brain-machine interfaces grow up.

Humans are inveterate usability engineers. Ever since an anonymous hunter wrapped a bit of hide around the handle of his spear to improve the grip, we’ve tried to make our tools easier to use—to make them feel more like extensions of our minds.

Brain Cap
Harsha Agashe, a PhD student in Jose Contreras-Vidal’s lab, wears the Brain Cap, a non-invasive, sensor-lined cap with neural interface software. (Source: John Consoli, University of Maryland)

Now, neuroscientists are trying to push that dream to its natural conclusion, building interfaces to connect our brains directly to machines. It’s not a new idea, but the field’s pace has accelerated so dramatically in the past few years that even insiders find it astonishing.

“If you’re looking at what’s on the cutting edge of neuroscience, they’re doing stuff that’s science fiction, it’s completely amazing,” says Tansy Brook, Communications Director for Neurosky in San Jose, CA.

While Neurosky has led the charge into the marketplace, building noninvasive systems that now power “mind-reading” toys and video games, basic researchers are pushing the boundaries even further. In the process, they’re confronting and conquering some of the toughest problems in neuroscience.

Put On Your Thinking Cap
Traditionally, scientists studying the brain have used small wire probes, sticking the electrodes directly into the brains of laboratory animals, and occasionally humans, to measure small signal changes in specific regions. That invasive approach has been tremendously informative, but even its proponents can rattle off a long list of limitations: the probes tend to stop working after a short period of time, they can become infected, and they generally measure only a small portion of overall brain activity.

Noninvasive interfaces based on the decades-old technique of electroencephalography (EEG) are also available, but have their own shortcomings. Or so it seemed.

NeuroSky CEO Stanley Yang with MindWave (Source: NeuroSky)

“When we entered this field back in 2005, we found there was a ... perception that [EEG] was not suitable for brain-machine interfaces involving multifunctional prosthetics,” says Jose Contreras-Vidal, PhD, Associate Professor of Kinesiology at the University of Maryland in College Park, MD. The thinking was that EEG didn’t offer sufficient resolution to distinguish, say, a leg extension from a leg rotation.

Since then, Contreras-Vidal and his colleagues have gradually tweaked traditional EEG to extract vastly greater amounts of information. The team discovered that covering the entire scalp with hundreds of electrodes, rather than the dozens used in most EEG studies, provides much higher resolution for movement signals. They also found that most previous EEG work had used the wrong demodulation techniques. The brain, it turns out, uses amplitude-modulated signals as well as frequency-modulated ones, and the signals extend into much higher frequency bands than anyone had checked before.

Taken together, the findings have allowed the researchers to build a noninvasive cap that allows a person to control a sophisticated prosthetic limb, at least in a controlled setting. The team is now collaborating with researchers and engineers at several other institutions and companies to get the system into clinical use, but Contreras-Vidal cautions that it won’t be trivial. “When you have a robot that closely interacts not only with the brain but the environment, you have to pay a lot of attention,” he says.

Playful Thoughts
If the brain is only interacting with a toy, however, the path to the market is smoother. That’s what engineers at NeuroSky discovered when they began introducing consumer-grade EEG systems. The company’s chips and sensors are the basis for the popular MindFlex toys from Mattel, and NeuroSky also sells its own lines of simplified EEG headsets and component parts for customers who want to assemble their own systems.

Flexible probe on plastic
A flexible probe printed on plastic could make invasive brain-machine interfaces more practical. (Source: George Malliaras)
NeuroSky is quick to point out that their devices are not substitutes for FDA-approved equipment. “I’m very careful to say that our technology is not a replacement for a medical or research-grade EEG system, it doesn’t have the same number of sensors, you’re not getting as much data,” says Brook.

Indeed, NeuroSky relies on a single sensing point on the forehead, and embeds the entire signal processing system inside a single chip. Device developers can tap into different streams of data from the chip. Besides the raw sensor readings, the system also provides processed signals representing alpha waves, or relaxation, and beta waves, or concentration. Toy and game makers with no previous EEG experience appreciate the simplicity of the processed readouts, while more advanced developers can apply their own algorithms to the raw signal.

While the single-sensor design is less flexible than traditional EEG, researchers are often surprised at the quality of the data these “toys” can produce. “People have purchased our technology and put it to use in a research environment almost on a lark … and they’ve actually had really good experiences with it,” says Brook. She adds that the low cost and user-friendliness has made EEG available to scientists who might not have used the technology otherwise.

Mass-market availability could also help ease public fears about future brain-machine interfaces. “You can buy it at Toys ‘R’ Us, and all of a sudden all of these kids know what an EEG is,” says Brook, adding that “we try to be good stewards of that.”

Getting Inside Your Head
Even proponents of noninvasive EEG-based approaches concede that their technology is unlikely to replace invasive probes for some applications. That’s because EEG can only sense signals that reach the scalp, while implanted probes can intercept messages in any part of the brain. For some types of interfaces, invasive systems will probably be the only solution—if researchers can overcome their limitations.

Highly flexible plastic probes
Highly flexible plastic probes, such as this one, could allow researchers to monitor brain signals at high resolution for a new generation of brain-machine interfaces. (Source: George Malliaras)

To be practical, implanted interfaces “need to record with high signal-to-noise ratio over a long period of time. This is a tall order that requires advanced materials,” says George Malliaras, PhD, Professor of Bioelectronics at the Ecole Nationale Supérieure des Mines de Saint Etienne in Gardanne, France.

Malliaras and his colleagues recently described a new probe design that might clear those hurdles. Instead of using the traditional gold wires, the researchers built a probe from an ultra-thin piece of plastic. Only 4 micrometers thick, the probe can conform closely to the folds that extend deep inside the cerebrum. Tests on rats showed that the design has a very high signal-to-noise ratio, and the investigators are now seeing how long the probes can continue operating without maintenance.

While the conformable probes are ideal for electrocorticography, which measures signals on the surface of the cerebral cortex, inserting the highly flexible pieces deeper inside the tissue would be tricky. To address that, Malliaras and his colleagues are developing stiff but soluble shuttles that could be stuck to the probes for insertion, then dissolved to leave the probes behind.

The team’s collaborators are already using the system to study epilepsy in rats, and Malliaras expects the new probes to reach humans eventually. “We will see more and more electrocorticography being performed in the clinic during brain surgery,” he says.Regardless of which approach they are taking, scientists in the field expect brain-machine interfaces to pay big dividends in both the clinic and the lab. As Contreras-Vidal explains, “a critical element of a successful [interface] is going to be plasticity of the brain. If we know how that’s done, then we have solved one of the grand fundamental questions in neuroscience, which is how the brain learns.”

This article was published in Bioscience Technology magazine: Vol. 35, No. 9, October, 2011, pp. 12-13.