Computers Obeying Brain Signals
Associated Press | April 4, 2005
By MALCOLM RITTER
ALBANY, N.Y. -- Researchers and volunteers around the world are taking early steps toward a complex but straightforward technological goal: to use electrical signals from the brain as instructions to computers and other machines, allowing paralyzed people to communicate, move around and control their environment literally without moving a muscle.
Most dramatically, that could help "locked-in" patients -- those who've lost all muscle movement because of conditions like Lou Gehrig's disease or brainstem strokes.
Take a look at what other people have accomplished lately with signals from their brains:
* A quadriplegic man in Massachusetts has shown he can change TV channels, turn room lights on and off, open and close a robotic hand and sort through messages in a mock e-mail program.
* Seven paralyzed patients near Stuttgart, Germany, have been surfing the Internet and writing letters to friends from their homes.
* At a lab in Switzerland, two healthy volunteers learned to steer a 2-inch, two-wheeled robot -- sort of like a tiny wheelchair -- through a dollhouse-sized floor plan.
And at labs in several universities, monkeys operate mechanical arms with just their brains. At the University of Pittsburgh, a monkey can feed itself chunks of zucchini and orange slices this way.
Research into harnessing brain signals goes back some 20 years. But lately it seems the research pot is starting to come to a boil, as advances in brain science, electronics and computer software have combined to push the field forward.
In fact, far more than half the scientific reports ever published in this area have appeared in the last three years alone, says researcher Dr. Jonathan Wolpaw. And while only about a half-dozen labs seriously worked in the field as late as the mid-1990s, now about 60 labs have gotten into it, he said.
"The field, in the last four or five years, has kind of exploded," he said.
To see firsthand what all the excitement is about, I signed on as an able-bodied research subject at Wolpaw's Brain-Computer Interface lab, part of the Wadsworth Center of the New York State Department of Health.
To somebody peeking into this little room, I'm just a middle-aged guy wearing a polka-dotted blue shower cap with a bundle of wires sticking out the top, relaxing in a recliner while staring at a computer screen.
But in my mind's eye, I'm a teenager sitting bolt upright on the black piano bench of my boyhood home, expertly pounding out the stirring opening chords of Chopin's Military Polonaise.
Not that I've ever actually played that well. But there's a little red box motoring across that computer screen, and I'm hoping my fantasy will change my brain waves just enough to make it rise and hit a target.
Some people have learned to hit such targets better than 90 percent of the time. During this, my first of 12 training sessions, I succeed 58 percent of the time.
But my targets are so big that I could have reached 50 percent by random chance alone.
Bottom line: Over the past half-hour, I've displayed just a bit more mental prowess than you'd expect from a bowl of Froot Loops.
That blue shower cap is actually stretchable nylon mesh, polka-dotted with 64 round white electrodes that eavesdrop on the electrical activity near the surface of my brain. They pass their measurements to a computer, which calculates the strength of one particular rhythm, called the "beta" rhythm. And the computer tells that little red box to either rise or fall, depending on how strong my beta rhythm is from moment to moment.
My job, then, is to learn to control the strength of my beta rhythm -- a body activity I didn't even know I had until a few weeks before walking into Wolpaw's lab.
I do know the beta rhythm is an "idling" rhythm, sort of like engine noise, with no particular function in normal life. It's coming from the portion of my brain that tells limbs to move and receives information related to movement. And it should get weaker when I imagine moving.
So on the first day Bill Sarnacki, the senior research technician who will guide me through the training, suggests that when the computer tells me to aim at the lower target I should let my mind go blank to make the little red box fall. When I'm supposed to aim at the upper target, I should imagine moving my hands to make the box rise.
Which is why my personal foray into neuroscience begins to the music of Chopin.
Before long I seek some advice from Scott Hamel, 44, of Averill Park, N.Y., who long ago mastered this task and moved on to tougher ones.
I'd watched him move that box vertically, horizontally and diagonally, by controlling two of his brain rhythms. He doesn't bother summoning up images any more, he said; "I just know how to make myself feel to make things happen."
On a good day, he said, "I can manipulate that thing around the screen almost like pushing something around a desk."
As for me, he suggests relaxing. "The harder I try, the worse I do," he said.
Some scientists envision taking the use of brain signals way beyond what's been done so far.
John Donoghue, chair of Brown University's neuroscience department and chief science officer of Cyberkinetics Neurotechnology Systems Inc. of Foxboro, Mass., talks about giving disabled people use of their arms and legs someday by using brain signals to drive their muscles.
Eventually, paralyzed people might even wear lightweight mechanical arms and legs that fit over their own limbs and would enable them to walk and reach for things, says Miguel Nicolelis of Duke University, who calls such devices "wearable robots." Nicolelis has done robot-arm work in monkeys and hopes to start studies in severely paralyzed people this year.
And Dr. Philip Kennedy of Neural Signals Inc. in Atlanta, who has tested brain sensors in seven locked-in patients since 1996, ponders the notion of helping such people speak someday. That would require planting electrodes in speech areas of the brain to give people control over 30 or so speech sounds, which would be produced by a synthesizer.
"It's not an insurmountable problem," Kennedy says.
That would be a huge jump from today's brain-controlled programs that can spell out words, but only a few letters per minute. The paralyzed patients near Stuttgart use such a program.
But even a relatively slow spelling device could make a huge difference to people with no good alternative to communicate, says Dr. Terry D. Heiman-Patterson. She is working with Wolpaw's laboratory in a project with her Lou Gehrig's disease patients at Drexel University.
That disease, formally called amyotrophic lateral sclerosis or ALS, gradually robs people of their ability to use their muscles. Eventually their breathing muscles stop working, and late-stage patients have to decide whether to go on a ventilator to stay alive.
"One of the reasons people choose to die over live is that they can no longer communicate," Heiman-Patterson said. "If we can unlock the ability to communicate with others...we may be able to change some of the choices people are making."
Even for people who can blink or direct their gaze to send signals, it may take 20 laborious minutes to ask to be taken to the bathroom or be turned over, she said "The difficulty becomes so great just to do that," she said, "that people say, 'I can't deal with this any more.'"
I do relax. In fact, it turns out I'm pretty good at making my mind go blank.
I imagine my brain is a chunk of cold white marble and resolutely refuse to think anything else for the four seconds the little red box takes to cross the screen. More often than not, it seems, the box sinks and hits the lower target.
Making that box rise, however, is a problem. Chopin works for a while, then seems to abandon me. I add an imagined jerk of my left wrist, which initially yanks the box to the top of the screen, but then loses its effectiveness too. Next I imagine both hands tickling the bottom of that danged box. Doesn't work.
Scooping up a hard-hit ground ball and throwing to first ... directing a Sousa march in the gazebo of a town square ... whacking a golf ball ... clawing at a dirt wall. (From Chopin to Sousa to clawing in dirt. Has it come to this?)
None of these works very well over my first few days of training, and my overall accuracy hasn't improved a lot either. I'm surprised when Wolpaw's colleague Dennis McFarland suggests a new strategy for the lower target: Imagine moving my feet. I thought the chunk-of-marble strategy was working well. But I tried his suggestion.
Gangbusters! Over the next three minutes I hit four out of every five targets overall, a startling improvement. The following run, I hit three out of every four, still gratifying. But my accuracy nosedives to around random-chance levels on the two runs after that. Before long, I go back to chunk-of-marble.
Eventually, I settle on the thought of waggling a baseball bat around for the upper target, and it seems to work pretty well. Between that and chunk-of-marble, I find myself enjoying occasional streaks of control, hitting two-thirds of the targets and sometimes much better.
I can't stay in the groove as long as I'd like. Usually, under my uncertain command, the red box flits across the screen like a butterfly buffeted by a summer breeze, its destination in doubt until the last instant.
But not when I'm at my best. I can make it glide upward like a party balloon or even jump like I'd punted it. And when I aim at the lower target, the box bumps its way downward, sometimes even dropping and running like a fumbled nickel.
What does that kind of control feel like?
Imagine you're a mediocre bowler. Imagine you've just released the ball, and in those long seconds as it approaches the pins it wanders toward the gutter and you're mentally telling it, "Get back! Get back there!
And now imagine that it does.
There might be an easier way to do this, if you're willing to have surgery.
When surgeons at Washington University in St. Louis, in cooperation with Wolpaw, placed tiny electrodes on the surface of the brains of four people recently, they achieved accuracies of 74 percent to 100 percent with just three to 24 minutes of training.
Some researchers put electrodes into the brain. Donoghue's Cyberkinetics system includes a chip about the size of a baby aspirin with 100 wire-like sensors, each thinner than a hair. The chip goes on the surface of the brain and the sensors extend about .04 inch below the surface.
Rather than monitor brain waves, the device intercepts a sample of the very signals that command arm movement, Donoghue said. So a patient doesn't have to learn how to control his brain waves, he just has to imagine moving his arm. "At that point," Donoghue said, "it works."
That's been the experience with the quadriplegic volunteer in Massachusetts, who showed he could move a cursor around a screen effectively, though less smoothly than healthy people can, Donoghue said. Cyberkinetics hopes to try its "BrainGate" system in four more patients this year and bring a product to market by 2007 or 2008.
Scientists who study implanted devices say scalp recordings like Wolpaw's just couldn't provide enough detailed information from the brain for elaborate control and natural movement of robotic arms or reanimated human limbs.
Researcher Andrew Schwartz at the University of Pittsburgh notes that his monkeys can move a cursor or robot arm in three dimensions, while Wolpaw's subjects can so far operate a cursor only in two dimensions. Schwartz also questions how consistently people can stay "in the zone" of peak performance with scalp recordings.
Wolpaw, for his part, says implanted electrodes don't pick up all the brain's signals for movement. It's like trying to play a symphony with only violins, he says. "You're using the violins alone to control the output," he said. "How well that will work remains to be seen."
What's more, he says, signals from implanted electrodes might be diminished over time by scar tissue, dying brain cells and slight displacements within the brain. As for staying in the zone, he said, that gets easier with practice. Right now, consistency is an issue with all the brain-signal approaches, he said.
He said he can't think of any task that shouldn't be achievable someday with scalp electrodes, in combination with some sophisticated software to handle the details. And while scalp electrodes haven't yet shown they can do everything implanted ones can, he said, they've already come pretty close.
"We may not have the same batting average," he said, but "we're playing in the same league."
I did not make Rookie of the Year.
"You're a success, you're just not a stellar success," Wolpaw told me. "You're at the lowest level we would call actual control."
That is, my accuracy had climbed to around 65 percent. About 80 percent of people reach or surpass that level within 10 sessions. Frankly, it felt more like influence than control.
But my 12 sessions were just an introduction, and I'd probably get better if I stuck with it, he said.
That doesn't necessarily mean I'd stumble across some magic visual image that would instantly transform me into a virtuoso. In fact, visual images are just training wheels, giving some guidance while the real learning happens.
What was really going on during the 2,592 times I shot at a target over a total of about six hours? My brain was learning through simple trial and error. Although my brain, Wolpaw said, was a bit slower on the uptake than most people's.
But the fact is, I was eventually able to make that red box sink fairly often without any need for imagery.
That may not sound like much. But in an area of brain science that's still in its infancy, few people on Earth can say they've accomplished that.