The simple act of opening an e-mail or saying “I love you” is out of reach for many people paralyzed by accident or disease, but 40 of them could get the chance to try out brain-computer interfaces (BCI) and reconnect with the world in two new pilot trials at Massachusetts General Hospital in Boston and the Wadsworth Center of the New York State Department of Health in Albany, New York.
Having previously tested their BrainGate implanted chip in four people in the first pilot study, scientists at Mass General and Brown University, in Providence, Rhode Island, announced on 10 June that they will recruit up to 15 additional participants for a trial of BrainGate2. This interface allows people who cannot move their arms to manipulate a computer cursor or a robotic arm with their thoughts. “Hopefully we will continue to learn quickly,” said Leigh Hochberg, who is leading the project along with John Donoghue at Brown. “We are really exploring the feasibility of using this device for people with tetraplegia.” Tetraplegia is paralysis of all four limbs.
And at the Wadsworth Center, Jonathan Wolpaw is gearing up to test his communication system in people with amyotrophic lateral sclerosis (ALS) at five Veterans Affairs (VA) hospitals across the country. VA personnel will oversee five beta testers at each location, and over a couple of years Wolpaw hopes to determine how useful the system is, and how an organization like the VA can support users.
Brain-computer interfaces connect the motor cortex with a computer or machine that interprets the neural signals and converts them into code that a computer or robotic limb can understand. BCIs allow people to pick out letters to communicate and move computer cursors. A person using a BCI may wear a cap with electrodes; other systems have the electrodes implanted beneath the skull. Scientists have already shown that monkeys can, via implanted electrodes, operate a robotic arm with their thoughts to feed themselves treats. (Velliste et al., 2008). The BrainGate trial was among the first to use implanted electrodes in people.
The field is burgeoning, and was featured on 60 Minutes last fall. When Jane Huggins, of the University of Michigan in Ann Arbor, started working on BCIs in the mid-1990s, “there were probably a handful of groups around the world,” she said. “Now, there are too many to count.” That’s a good thing, Huggins said: “We have no shortage of problems to be solved.”
The solutions will involve interdisciplinary collaboration, Donoghue said. Developing BCIs requires neuroscientists, engineers, computer scientists, and mathematicians to collect, decipher, and apply the brain’s electrical signals to control technology.
That technology is designed to give an artificial interface to people who have lost their natural ability to move or speak. For a person who suffered spinal cord injury leading to tetraplegia, that could mean something as simple as reaching out to grab a cup of water. “I just want to scratch my nose,” one person told Donoghue.
For someone who had a stroke or who has late-stage ALS, even communication may be gone, essentially “locking in” an active mind. “The number one priority for people is usually communication,” Huggins said. Some current communications systems rely on the blink of an eye or the twitch of an eyebrow—but those muscles can tire, and for some people, even those movements are impossible.
I’m Thinking of a Letter Between A and Z…
Neuroscientist Scott Mackler of the University of Pennsylvania, who has ALS, has used Wolpaw’s communication system for more than three years; it allows him to supervise his laboratory’s research on cocaine addiction. “BCI has made me independent,” he wrote in an e-mail to ARF, selecting each letter individually using the BCI.
The system Mackler uses to pick out his words letter by letter relies on the P300 response, short for “positive deflection of brain activity peaking 300 milliseconds after a stimulus.” The user, wearing an electrode cap that is hooked up to an electroencephalograph (EEG), faces a computer screen with a grid of letters and numbers (see picture). To spell a word, the person focuses on the first letter desired. The program flashes different letters, in groups or individually, and every time the chosen character lights up, the brain provides a little response—“Aha! That’s my letter!” By integrating these flashes with the EEG response, the computer program can figure out what letter the person is after. The system works as a surrogate keyboard on any Windows program.
“I was amazed how easy it worked,” Mackler wrote. In April, using a similar system, graduate student Adam Wilson at the University of Wisconsin in Madison was able to post to the social micro-blogging website Twitter. (See video.)
The P300 communication system
A person with ALS (right) wears an EEG cap. The BCI system detects his brain's response when the letter he wants lights up (left). Image credit: Jonathan Wolpaw, Wadsworth Center View larger image
The approach used by Donoghue and Hochberg taps neural signals to produce one-, two- and three-dimensional motion. The BrainGate electrode array reads electrical signals from dozens of individual neurons in the motor cortex. Surgeons insert the array into the distinct “knob” of the motor cortex that normally controls arm movement, so that the electrodes can eavesdrop on the neural impulses of 96 neurons in that area. “That motor cortex still has the ability and the desire to move those muscles,” Hochberg said. The BCI must first be calibrated for the person using it; the user imagines specific motions so the computer can “learn” to match electrical impulses to intended action. Then, the computer should be able to read and carry out new instructions. The motor cortex has far more than 96 cells, of course, but by listening in on those few, the computer can extrapolate the desired action.
In the first preclinical phase of the BrainGate trials, the researchers implanted their device into two people whose spinal cords were injured, one person who had a brainstem stroke, and one with ALS. With their brains literally plugged in to the system via a connector atop their heads, the volunteers were able to control a cursor on a computer screen.
The first person tried the system three years after an injury had paralyzed him from the neck down, but his motor cortex was still ready to go. “We saw the cells in the motor cortex fire immediately,” Hochberg said. “That was, for me, the hallmark scientific moment.” With BrainGate, the man was able to open e-mail, play the video game Pong, open and close a prosthetic hand, and maneuver a robotic arm (Hochberg et al., 2006).
Ultimately, researchers want to go well beyond simple typing and point-and-click applications. Someday, Donoghue envisions, a person with a severed spinal cord could still hit the basketball court, relying on technology that bypasses the spinal cord to allow the brain to communicate with the muscles.
With respect to ALS, a good BCI could even alter choices about life and death, said Philip Kennedy, chief scientist and CEO of Neural Signals, Inc., in Duluth, Georgia. Kennedy is working to develop a mind-to-spoken-word translator for locked-in patients. Many people with ALS, when they can no longer breathe on their own, opt not to go on a ventilator. Mackler, in fact, originally made that choice. Part of this decision may have to do with people fearing that they’ll be trapped in their body with no way to interact with friends and loved ones. BCIs can partly restore a person’s social interface and, Kennedy claimed, may make life more worth living to some.
Eavesdropping on the Neural Symphony
If you were to record a choir, where would you place the microphones? On individual singers—say, the first soprano and third tenor? Or would you position the mike near groups, such as the altos and baritones? Perhaps you might go farther out, recording the group as a whole from a seat in the audience.
This is a question BCI scientists face in deciding where to put the electrodes, Huggins said. The simplest solution is a surface EEG, but the skull muffles and distorts the signals from inside. The system Donoghue and Hochberg use is an implanted electrode array the size of a baby aspirin. It allows the scientists to wiretap individual neurons, but it requires brain surgery to implant and can’t be easily tweaked or repaired. A middle ground is electrocorticography (ECoG), which uses electrodes placed below the skull and dura mater but atop the brain. Those electrodes can be bigger, reaching more areas than the BrainGate electrodes, since they aren’t inserted into the brain, although surgery is still required. Each system has its proponents.
Huggins is most interested in EEG as the fastest route to a broadly valuable device. “There would be a great advantage to brain-computer interface work in getting something quickly,” she said. EEG systems have the potential now, she said, to help people communicate, and are nearly ready for home use.
In contrast, Hochberg argued that implanted electrodes are the way to go. “If you want to listen to the symphony, you’d rather be in the concert hall, not across the street,” he said. “The signals that are most immediately related to the onset of movement are those that can be recorded intracortically.”
For his part, Wolpaw isn’t convinced that deeper is better for BCIs. “In the long term, I think ECoG may prove the best option,” he said. Wolpaw described ECoG as “super EEG,” resulting in a cleaner, higher-amplitude signal because it reads brain activity beneath the skull. These larger devices, perhaps four millimeters in diameter, can also reach broader areas of the cortex than implanted arrays.
Out of the Lab and Into the Living Room
Ultimately, the best solution will simply prove to be the one that works. “The most important technology is the one that works reliably and restores function,” Hochberg said. Alas, there are plenty of hurdles between the current devices and fully practical BCIs. One barrier to moving BCI out of the lab and into the living room, Huggins noted, is how a user will switch the BCI on and off. For example, your computer will go to sleep if you don’t touch it for a while, but with a BCI, “you’re always touching it,” she said. Since some BCIs rely on the person visualizing movement, it’s possible the machine might not always correctly distinguish between imagined and intended actions. You might think about punching your boss in the nose, but you wouldn’t want the BCI to actually tell a mechanical arm to do it. The solution will likely be some sort of “pause” mode, which the user might turn on to halt any BCI-based movement while, say, looking out the window or watching an action movie. A specific thought, or sequence of thoughts, could flip the device on and off, but Huggins suspects that eventually, a person might think that particular “code” unintentionally.
Another problem is that current BCIs give highly variable results from day to day. In medical lingo, they could even be described as “ataxic,” Wolpaw wrote in a 2007 review. A user might move the cursor to the right target nearly every time in one lab session, but struggle in the next. Part of the reason, Wolpaw said, is that BCIs are, by definition, unnatural devices that translate signals meant for the muscles into artificial action. Normally, movement is coordinated by several parts of the nervous system, including the cortex and spinal cord, not just the small area scientists can tap with electrodes. In addition, natural motion depends on constant adaptation of the system—something machines are ill equipped to deal with.
In the real world, this erratic performance would be frustrating, even dangerous, if a caretaker were not nearby to take over for the machine. Because of these uncertainties, many researchers shy away from applications such as wheelchairs. “The failure modes are too scary,” Huggins said.
Wolpaw’s trial, in collaboration with the VA, will address some basic questions about how BCIs fit into daily routine. How much will people actually use the P300 BCI? Will it improve their quality of life? What kind of tech support will be necessary? For the system to be truly practical, it must be easy to use without too much assistance.
The BrainGate team is also working on making their device more practical with a system that is fully implanted beneath the skull. “The external hardware is something everybody knows has to disappear,” Hochberg said. Having machinery cross the skin invites infection, and this risk would be eliminated with a wireless transmitter under the skin. Hochberg hopes to have such a system in a few years. Moreover, to be practical for a person who needs assistance, the device must last for years without repair.
Another question is how a BCI will fit into the life of someone whose nervous system is changing, such as a person with ALS. The disease can affect the brain as well as the spinal cord, and is sometimes associated with frontotemporal lobar degeneration. Doctors would not want to hook a BCI into neurons that are unlikely to last. Robert Welsh, also at the University of Michigan, recently began a study using functional magnetic resonance imaging to find out what changes occur in the brain of someone with ALS. “We do not know how person A who has ALS, and person B who has ALS, might progress differently,” he said, adding that understanding the variation between people and over time will be essential to the long-term success of implanted BCIs.
At the same time, BCI’s recording from changing neurons might offer researchers new insight. BCIs offer an unprecedented window into the mind’s inner workings. “This is the first chance I know of to record, for months or years on end, from dozens or perhaps a hundred or more neurons in the human brain,” Hochberg said. Data from BCI studies will tell scientists how the brain—healthy or otherwise—adapts, and give a more detailed picture of which neurons perform different functions. Describing how the brain changes in ALS or other diseases might help scientists who are trying to develop treatments, Hochberg noted, This means that aside from the potential applications for BCI, there is also basic science knowledge to be gained, and that makes the studies doubly worthwhile, Kennedy said. “Even if we fail, we sort of win.”—Amber Dance.
- Velliste M, Perel S, Spalding MC, Whitford AS, Schwartz AB. Cortical control of a prosthetic arm for self-feeding. Nature. 2008 Jun 19;453(7198):1098-101. PubMed.
- Hochberg LR, Serruya MD, Friehs GM, Mukand JA, Saleh M, Caplan AH, Branner A, Chen D, Penn RD, Donoghue JP. Neuronal ensemble control of prosthetic devices by a human with tetraplegia. Nature. 2006 Jul 13;442(7099):164-71. PubMed.
- Wolpaw JR. Brain-computer interfaces as new brain output pathways. J Physiol. 2007 Mar 15;579(Pt 3):613-9. PubMed.
- Takahashi M, Tomizawa K, Ishiguro K. Distribution of tau protein kinase I/glycogen synthase kinase-3beta, phosphatases 2A and 2B, and phosphorylated tau in the developing rat brain. Brain Res. 2000 Feb 28;857(1-2):193-206. PubMed.
- Takashima A, Murayama M, Yasutake K, Takahashi H, Yokoyama M, Ishiguro K. Involvement of cyclin dependent kinase5 activator p25 on tau phosphorylation in mouse brain. Neurosci Lett. 2001 Jun 22;306(1-2):37-40. PubMed.
- Moritz CT, Perlmutter SI, Fetz EE. Direct control of paralysed muscles by cortical neurons. Nature. 2008 Dec 4;456(7222):639-42. PubMed.
- Birbaumer N, Cohen LG. Brain-computer interfaces: communication and restoration of movement in paralysis. J Physiol. 2007 Mar 15;579(Pt 3):621-36. PubMed.
- Hatsopoulos NG, Donoghue JP. The science of neural interface systems. Annu Rev Neurosci. 2009;32:249-66. PubMed.
- Donoghue JP. Bridging the brain to the world: a perspective on neural interface systems. Neuron. 2008 Nov 6;60(3):511-21. PubMed.