“He had high blood pressure and didn’t take his medicine,” Kennedy says. “It was a catastrophic stroke of the brain stem.”
In a coma for weeks, Ray finally woke up in the Atlanta VA Medical Center, his mind intact but his body unable to move or communicate except by the slightest quiver of a few muscles in his face, including his eyelids. He was what doctors call “locked-in.” Blinking twice for yes and once for no, he agreed to participate in Kennedy’s study.
Kennedy and Emory neurosurgeon Roy Bakay implanted a Neurotrophic Electrode near the part of Ray’s brain that controlled his left hand. The outer end was attached to an amplifier and radio transmitter on his skull under the scalp. In the months that followed, Kennedy encouraged Ray to think about moving a computer mouse with his hand.
As Ray imagined moving the mouse, there was an increase in electrical activity among the neurons that would have controlled that action if his hand could move. These brain impulses were transmitted to a receiver on his pillow, where they were deciphered and translated into digital commands sent to a nearby computer. Over time, the computer began to obey Ray’s neural signals. Within six months Ray was moving a cursor on the screen through intention alone, communicating by clicking on icons for phrases like “I’m cold.” Despite a host of excruciating health problems, including infections (not related to the implant), Ray kept working with the researchers, although it clearly exhausted him. After more months of practice, he could spell words and hold brief conversations.
When Kennedy asked what he felt when he moved the cursor, Ray spelled “NOTHING.” It was a surprising, significant moment. Ray had learned to control the cursor without thinking about moving his paralyzed arm. Neural activity that had been linked to arm and hand movement had changed. Now his brain was communicating directly with the computer.
“The brain is very adaptable,” Kennedy says in the Irish accent he has kept after almost three decades in the United States. “The brain’s plasticity is the key thing to this whole field.”
The FDA doesn’t readily approve radical procedures like brain implants, so most of Kennedy’s colleagues have pushed the envelope not in humans but in other primates—for instance, a tiny owl monkey named Belle. In 2000 Duke University neuroscientist Miguel Nicolelis trained Belle to move a joystick in sync with a light. If she did the task correctly, she was rewarded with juice.
Belle wore a hat attached to about 100 wires as fine as human hairs. The wires were implanted in Belle’s motor cortex, the part of the brain that plans and initiates movement. As she moved the joystick, her neural signals were picked up by the wires and sent to a computer in the next room. The computer sent them on to a robotic arm, which precisely mimicked the action of Belle’s arm. At the same time, Nicolelis transmitted the brain signals over the Internet to the Touch Lab at MIT in Cambridge, Massachusetts. There, 700 miles from Duke, the neural commands operated another robotic arm. The virtual and physical world had merged.
In January 2008, working with the Computational Brain Project of the Japan Science and Technology Agency, Nicolelis took another step forward, this time with Idoya, a rhesus monkey trained to walk upright on a treadmill for treats. Electrodes were implanted in the part of Idoya’s brain that controls leg movement; these devices recorded the activity of 250 to 300 neurons that fired when her ankles, knees, hip joints, and feet moved or were about to move. Fluorescent stage makeup allowed a special high-speed video camera to capture the details of her limb motions.
The video and neural signals were then combined to show which muscle movements resulted from which neuron firings and how the activity of the neurons appeared to anticipate movement. In the end, computer analysis predicted Idoya’s leg movements about a second before the animal actually carried them out. As if all this weren’t enough, the system transmitted the predictions over a high-speed Internet connection to Kyoto, Japan, and into the actuators of a robot named CB-1 (for Computational Brain), which was designed to have a remarkably humanlike range of motion. Idoya walked and so did CB-1, in perfect sync.
The 500-pound, 5-foot-tall robot was much larger than the 12-pound, 32-inch-tall monkey whose neural signals were directing it, and this underscored a simple yet remarkable point: Implant technology could enable brainpower to control a huge object (like a robot crane) or a tiny one (like a microscopic surgical tool) just as easily as a life-size mechanical arm.
For the technology to truly aid humans, robot limbs will have to function in a lifelike way. That’s where neuroscientist Andrew Schwartz of the University of Pittsburgh School of Medicine comes in. He recently implanted an array of electrodes about the size of a freckle in the motor cortex of two macaques. The animals’ arms were gently restrained and a mechanical arm with a grasping claw was strapped to their left sides. The neural signals captured by the implant had been produced by the macaques’ brains to direct their arm movements. Instead the signals were shunted to a computer, which in turn directed the movements of the robotic arm. The monkeys didn’t have to think about moving the robotic arm; they simply reached out fluidly with the prosthetic limb as if it were part of them. The success of this experiment was due in part to a computer capable of rapidly interpreting brain signals based on the monkeys’ desire to move their limbs. This “advance warning system” moved the robotic arm in just 150 milliseconds, about the length of time it takes brain activity to spur real arm movement. The rapid response helped the monkeys use the robotic arm in a natural way, reacting quickly if they were about to drop a piece of food, for instance, and refining movements in real time.
Work like this lends credence to the cyborg-maker’s long-sought goal: the possibility, in the not-too-distant future, of helping the paralyzed walk, reach, and grasp. Front and center in this effort is Northwestern University neuroscientist Lee Miller, who injects local anesthetic into a monkey’s arm so that the limb is temporarily paralyzed. Then, instead of sending neural signals from the animal’s brain to a robot, he shunts them back into the muscles of the paralyzed arm, thereby bypassing the spinal cord. “The signals are going to a stimulator that is electrically stimulating those same muscles,” Miller explains. “So essentially it allows the monkey to use his arm again, flexing the wrist and playing a video game all entirely voluntarily, despite the fact that the arm is actually paralyzed.”
Merging man and machine
The spectacular successes of brain implants in primates has paved the way for new human trials, including one at Brown University, where neuroscientist John Donoghue is moving ahead with BrainGate, a minuscule array of tiny, spikelike electrodes implanted in the motor cortex. Candidates are quadriplegics, with all four limbs paralyzed due to ALS, spinal cord injury, or brain stem stroke. So far, three patients implanted with BrainGate can voluntarily modulate several dozen neurons sufficiently to type on a screen, move a prosthetic hand, or control a robotic arm.
“Our goal is to help restore communication and independence,” says Donoghue’s colleague Leigh Hochberg, a neuroscientist with the Department of Veterans Affairs. One patient, a 37-year-old with ALS, died 10 months into the trial after his respirator was inadvertently disconnected. His untimely death, and the progress he made while participating in the experiments, were especially moving to Hochberg. “He could demonstrate that his mental status was fully intact,” Hochberg says. “He had great insight into his disease and the research we were doing and great humor as well.”
Even with the limited number of subjects, the human research has already confirmed the monkey findings and answered important questions about how the brain works. “One thing we wondered was how a particular part of the brain functioned years after an arm and hand hadn’t moved due to disease or injury,” Hochberg says. “We found some insight thanks to one of our first participants, who had a spinal cord injury. He was paralyzed, but the moment he thought about using his hand, we saw changes in neural activity in the specific part of the motor cortex associated with hand movement. Different neurons fired at different rates depending on what he imagined performing.”