Matrix realized: connecting brains to computers could circumvent disabilities.

At the University of Tubingen in Germany, neurobiologist Andrea Kubler works with a 49-year-old patient whom she identifies only as H.S. Like many of Kubler's patients, H.S. suffers from amyotrophic lateral sclerosis (ALS), a degenerative disease that slowly breaks down the nerve cells necessary for motion. The disease has paralyzed H.S., stripping him of the motor functions that most people take for granted: sitting up, eating, and even breathing.

H.S. was diagnosed with ALS 14 years ago, and a permanent ventilator has done his breathing for him since 1993. From then until recently, H.S. could converse with Kubler and his caregivers only by blinking his eyes.

Many people paralyzed by a variety of causes--strokes, tumors, or traumatic brain injuries, to name a few examples--communicate rather well by using their remaining, although impoverished, muscle power. For example, French editor Jean-Dominique Bauby, who was paralyzed by a stroke, dictated a 132-page novel (The Diving Bell and the Butterfly, 1997, Knopf) just by blinking his left eye. Similarly, decades after physicist Stephen Hawking was paralyzed by ALS, he continues to give world-renowned lectures by feeding words into a computer program with a flick of his finger.

Eventually, H.S.'s paralysis might prevent him from even blinking his eyes to express himself. He'd then be in a state that doctors aptly refer to as "locked-in syndrome." Thousands of paralyzed people worldwide are in that socially closed-off state.

"We know that communication is very important to maintaining quality of life," says Kubler.

A device developed a few years ago by Kubler and her colleagues gives him the option of communicating by moving a cursor on a screen using sheer brainpower. Other scientists are developing neurological tools that may eventually enable patients such as H.S to flex mechanical limbs, steer a motorized wheelchair, or operate robots that respond to their needs, much as specially trained dogs assist some less-incapacitated patients.

These preliminary brain-computer interfaces, or BCIs, are machines that could link patients' fully functioning brains to the outside world.

BCIs "could change the picture as to what you're going to tell people as they get disabled," says John Wolpaw, a neurologist at the Wadsworth Center of the New York State Department of Health in Albany. "The fact that you're paralyzed doesn't have to mean a lousy quality of life.

INS OF OUTPUT Over the past 3 decades, scientists working on BCIs have produced a vast amount of research, some currently applied to patients but most still in the experimental realm (SN: 8/28/99, p. 142). All the resulting devices fall neatly into one of two categories: those that direct electric flow into the brain and those that pick up the brain's electric output.

Several popular therapeutic devices fall into the first group. For example, by converting sound waves into electrical pulses that stimulate the auditory nerve, cochlear implants improve the hearing of more than 50,000 deaf or hearing-impaired people in the United States.

The second BCI category, devices known as neural prostheses, has proved far trickier. Delivering electric stimulation can make do with "cruder technology" than detecting it does, says Andrew Schwartz, a neural physiologist at the University of Pittsburgh. Tapping into the brain's electric signals requires electrodes that are far smaller and more precise than those used to stimulate neurons. Beyond that, scientists have had trouble deciphering whatever brain signals they do pick up.

However, Bill Heetderks of the National Institute of Biomedical Imaging and Bioengineering in Bethesda, Md., says that he and other neuroscientists working on the problem in the late 1980s had few doubts that such devices would eventually succeed. Researchers had long been making crude recordings of neural signals with electrodes implanted in the brains of animals or stuck to the scalps of people.

Although some hints of the brain's inner signals radiate to the scalp in the form of electroencephalogram (EEG) waves, these indirect signals are far weaker than the electrical pulses running between brain cells themselves. Therefore, to decipher movements as complicated as throwing a softball or reaching for a snack, many scientists went straight to the source: individual neurons inside the brain.

"If you take signals directly from a large number of neurons, you can get a very good, detailed estimate for how an arm is going to move in space," says Heetderks.

MIND OVER MATTER Researchers originally gathered signals by poking single neurons with bulky needles. But today, an electrode about the diameter of a baby aspirin can hold more than 100 tiny needles, each thinner than the diameter of a human hair. With the needles jutting out from their platform like a tiny bed of nails, each electrode has the potential to record from many neurons at the same time.

For Richard Andersen, a neuroscientist at California Institute of Technology in Pasadena, these electrodes are a valuable resource for collecting information on movement. Over the past 6 years, he and his colleagues have used such electrodes to record the brain activity of rhesus monkeys.

Andersen's group focuses on the posterior parietal cortex and the premotor cortex, two brain areas that researchers suspect plan a person's motions before the body actually moves. By recording impulses in those areas, says Andersen, a researcher can in effect decipher the brain's intentions. Then, with the right computer software, investigators can devise machines that will act out the intentions of a patient who's unable to move.

"You could do almost anything with this [type of device]" says Andersen. "You could implant it in speech regions, and when the subject forms words, it could operate a voice synthesizer."

Andersen adds that electrodes might be placed in locations that would enable paralyzed patients to communicate their deepest feelings, such as being pleased to see a visitor, without saying a word. "We could read out their emotions," he says.

Other researchers, including Schwartz, are using similar electrodes to teach monkeys to operate a mechanical arm. Miguel Nicolelis of Duke University in Durham, N.C., has implanted electrodes in a monkey's motion-generating motor cortex. The monkey's brain impulses, in response to an animated game on a computer screen, swing a mechanical arm located in another room.

In contrast, each of Schwartz' monkeys sits next to a motorized arm that's being operated by the animal's brain. With their own arms restrained, the monkeys use the mechanical appendages to reach for food.

Each approach uses a computer program with algorithms that translate a cacophony of neural impulses into a clear signal that gives the arm directions. For example, Schwartz and his colleagues devised a program based on the observation that each neuron in the motor cortex has a preferred direction--it fires more frequently when the monkey wants to move its real arm in a particular direction. As the monkey gestures in a circle, different groups of neurons pulse to direct movement at each point along the circle's rim.

With that information, directing the neural signals to operate the arm is "really pretty simple," says Schwartz. "All you have to do is figure out what direction each [neuron] is related to, and then you can come up with an algorithm that basically takes a vote [among the different neural signals] and tells you what direction you're moving."

CYBORG SPELLER As Schwartz, Nicolelis, and Andersen do, most researchers test their prototype neural prosthetics in laboratory animals such as monkeys or rats. However, two scientists have implanted prosthetic BCIs in people.

In 1996, neurologist Philip Kennedy of Neural Signals, a company based in Atlanta, and his colleagues were the first group to implant a neural electrode into a person.

Kennedy's group doesn't use the beds of needles employed by most researchers in this area, but instead works with a radically different electrode consisting of a tiny glass cone about the size of a ballpoint-pen tip. Within the cone sit three electricity-conducting wires and a bit of neural growth factor, a protein that encourages neurons to extend into the electrode.

"I decided that it was better to bring the brain into an electrode rather than the other way around," Kennedy says.

The device is enclosed within the skull and transmits signals to external receivers. The first patient that Kennedy fitted with the electrode provided little more than sample readings before she succumbed to her illness. However, the second patient, whom Kennedy refers to as J.R., gave the researchers results beyond their wildest expectations. A locked-in patient paralyzed by stroke, J.R. eventually learned to control a cursor on a computer screen, spelling out words using only brain power.

"J.R. was our first cyborg," Kennedy muses. "He had a spelling rate of about 3 characters per minute. It's not terribly fast, but it's better than nothing." J.R. died in 2002, but Kennedy is currently testing the approach on other patients.

Last May, John Donoghue and his colleagues at Cyberkinetics in Foxboro, Mass., became the second team to implant neural electrodes in people. The company's BrainGate system uses multiple-needle electrodes connected directly to a computer. Within 2 months, a 25-year-old quadriplegic hooked up to BrainGate could use the system to open e-mail, change television channels, switch lights on and off, and operate a robotic arm.

Although the patient lost the use of his arms 3 years ago, the brain areas controlling his hands and arms still function normally. "When we asked [the patient] to imagine moving his hands around in space, the cursor moved. He was able to do the kinds of point-and-click actions that someone would do with a mouse, but we bypassed his hand and went straight to the computer," says Donoghue. Cyberkinetics plans to market the BrainGate system by 2007.

FROM THE TOP Some researchers are concerned that BrainGate's open design, with wires extending through the skin from the brain-implanted electrodes, could leave a person vulnerable to infection. However, even a closed system such as Kennedy's, where an implant stays sealed under the skull and transmits information by telemetry, poses some problems.

Researchers insert the electrodes in a "kind of a shotgun approach," says Andersen. If the needles don't end up next to a cell, they don't pick up any signal.

Additionally, he says, natural movement in the brain and scarring around the tips of the electrodes sometimes dampen the electrical signals available to researchers.

Some scientists predict that the solution to these problems lies not in improving implantable electrodes, but in developing new systems to record brain waves from outside the skull's surface, the same pulses typically observed in an EEG. Although many researchers expect that these signals would be too weak and nonspecific to control BCIs, "that's an unproved assumption that people engaged in [competing] studies have been portraying for years" says Wolpaw.

In the Dec. 21, 2004 Proceedings of the National Academy of Sciences, Wolpaw and his colleagues published research showing that with improved EEG methods, four patients with different amounts of paralysis could each operate a cursor on a screen smoothly and quickly without undergoing surgery. By learning to vary the intensity of their brain waves, each subject has learned to move the cursor in any direction at various speeds. Kubler is using a similar approach with H.S. and other patients.

Donoghue suggests that different methods will be required to meet the wide variety of patient needs. The final hurdle for some techniques, he speculates, may be getting people to feel less squeamish about using devices that attach directly to the brain.

"This is really the dawn of neurotechnology," Donoghue says. "Someday, we'll be nearly as comfortable about putting devices in our brains as we [are about those that go] in our hearts, stomachs, or teeth. It could have huge effects."