To See My Life Dream
Just Read...

Brain implant turns thought to robot acts
(Researchers train a computer to read monkeys' intentions directly from their neural activity)

RICK WEISS, Of The Oregonian
10/13/03

"It's a major advance," University of Washington neuroscientist Eberhard Fetz said of the monkey studies. "This bodes well for the success of brain-machine interfaces."

The experiments, led by Miguel Nicolelis of Duke University in Durham and published Monday in the journal PLoS Biology, are the latest in a progression of increasingly science fiction-like studies in which animals -- and in a few cases people -- have learned to use the brain's subtle electrical signals to operate simple devices.

Until now, those achievements have been limited to "virtual" actions, such as making a cursor move across a computer screen, or to small two-dimensional actions such as flipping a little lever that is wired to the brain.

The new work is the first in which any animal has learned to use its brain to move a robotic device in all directions in space and to perform a mixture of interrelated movements -- such as reaching toward an object, grasping it and adjusting the grip strength depending on how heavy the object is.

"This is where you want to be," said Karen Moxon, a professor of biomedical engineering at Drexel University in Philadelphia. "It's one thing to be able to communicate with a video screen. But to move something in the physical world is a real technological feat. And Nicolelis has taken this work to a new level by quantifying the neuroscience behind it."

The experiment The device relies on tiny electrodes, each one resembling a wire thinner than a human hair. After removing patches of skull from two monkeys to expose the outer surface of their brains, Nicolelis and his colleagues stuck 96 of those tiny wires about a millimeter deep in one monkey's brain and 320 of them in the other animal's brain.

The surgeries were painstaking, taking about 10 hours, and ended with the pouring of a substance like dental cement over the area to substitute for the missing portions of skull. The monkeys were unaffected by the surgery, Nicolelis said. But now they had tufts of wires protruding from their heads, which could be hooked up to other wires that ran through a computer and on to a large mechanical arm.

Then came the training, with the monkeys first learning to move the robot arm with a joystick. The arm was kept in a separate room -- "If you put a 50-kilogram robot in front of them, they get very nervous," Nicolelis said -- but the monkeys could track their progress by watching a schematic representation of the arm and its motions on a video screen.

The monkeys quickly learned how to use the joystick to make the arm reach and grasp for objects, and how to adjust their grip on the joystick to vary the robotic hand's grip strength. They could see on the monitor when they missed their target or dropped it from having too light a grip, and they were rewarded with sips of juice when they performed their tasks successfully.

While the monkeys trained, a computer tracked the patterns of bioelectrical activity in the animals' brains. The computer figured out that certain patterns amounted to a command to "reach." Others, it became clear, meant "grasp." Gradually, the computer learned to "read" the monkeys' minds.

Then the researchers did something radical: They unplugged the joystick so the robotic arm's movements depended completely on a monkey's brain activity. In effect, the computer that had been studying the animal's neural firing patterns was now serving as an interpreter, decoding the brain signals according to what it had learned from the joystick games and sending the appropriate instructions to the mechanical arm.

An amazing result At first, Nicolelis said, the monkey kept moving the joystick, not realizing her own brain was now solely in charge of the arm's movements. Then, he said, an amazing thing happened.

"We're looking, and she stops moving her arm," he said, "but the cursor keeps playing the game and the robot arm is moving around."

The animal was controlling the robot with its thoughts.

"We couldn't speak. It was dead silence," Nicolelis said. "No one wanted to verbalize what was happening. And she continued to do that for almost an hour."

At first, the animals' performance declined compared to the sessions on the joystick. But after just a day or so, the control was so smooth it seemed the animals had accepted the mechanical arm as their own.

"It's quite plausible that the perception is you're extended into the robot arm, or the arm is an extension of you," agreed the University of Washington's Fetz, a pioneer in the field of brain-controlled devices.

Possibilities for the paralyzed John Donoghue, a neuroscientist at Brown University developing a similar system, said paralyzed patients could be the first to benefit by gaining an ability to type and communicate on the Web, but the list of potential applications is endless, he said. The devices might allow quadriplegics to move their own limbs again by sending signals from the brain to various muscles, leaping over the severed nerves that caused their paralysis.

"Once you have an output signal out of the brain that you can interpret, the possibilities of what you can do with those signals are immense," said Donoghue, who recently co-founded a company, Cyberkinetics Inc. of Foxboro, Mass., to capitalize on the technology.

Both he and Nicolelis hope to get permission from the Food and Drug Administration to begin experiments in people next year. Nicolelis also is developing a system that would transmit signals from each of the hundreds of brain electrodes to a portable receiver, so his monkeys -- or human subjects -- could be free of external wires and move around while they turn their thoughts into mechanical actions.

"It's like multiple cellular phone lines," Nicolelis said. "As my mother said, 'You can dial your brain now.' "