A monkey sits on a bench, wires running from its head and wrist into a small box of electronics. At first the wrist lies limp, but within 10 minutes the monkey begins to flex its muscles and move its hand from side to side. The movements are clumsy, but they are enough to justify a rewarding slug of juice. After all, it shouldn't be able to move its wrist at all.
A nerve connection in the monkey's upper arm had previously been blocked with an anaesthetic that prevented signals travelling from its brain to its wrist, leaving the muscles temporarily paralysed. The monkey was only able to move its arm because the wires and the black box bypassed the broken link.
The monkey was in Eberhard Fetz's lab at the University of Washington in Seattle. The experiment, performed last year, was the first demonstration of a new treatment that might one day cure paralysis, which is typically caused by a broken connection in the spinal cord. Though much work has focused on using stem cells to regrow damaged nerve fibres, some researchers believe that an electronic bypass like this is equally viable.
The idea is to implant electronic chips in the relevant regions of the brain to record neural activity. Then a decoder deciphers the neural chatter, often from thousands of neurons, to figure out what the brain wants the body to do. These messages must then be relayed - ideally wirelessly - to electrodes that deliver a pulse of electricity to stimulate the muscles into action. Such "brain chips" are already restoring hearing to the deaf and vision to the blind, and helping to stave off epileptic fits, so the idea isn't as far-fetched as it might sound (see "Bionic medicine").
Every step of progress in tackling paralysis has been hard won. One of the early demonstrations that it may be possible emerged in 2003, when José Carmena, then at Duke University in Durham, North Carolina, successfully created an interface between brain and machine that allowed his lab monkeys to play a computer game using only their minds.
To gain a juice reward, the monkeys had to move a cursor - initially with a joystick - to hit a target on the computer screen. Beforehand, Carmena and his colleagues had implanted several chips throughout the parietal and frontal lobes of the monkeys' brains - regions known to plan and control movement. Each chip held up to 64 electrodes, which recorded the firing of the surrounding neurons as the monkeys manipulated the joystick.
Once the system had successfully decoded the chatter from the monkeys' neurons, the program stopped responding to the joystick's movement altogether and relied solely on the monkeys' thoughts to control the cursor. Eventually even the animals worked this out and stopped holding the joysticks as they completed the task (PLoS Biology, vol 1, p 42).
Manipulating a cursor on a computer screen is one thing, but whether such brain chips could translate the more complicated tasks of daily life remained an open question until 2004, when John Donoghue and colleagues from Cyberkinetics in Providence, Rhode Island, implanted a 100-electrode chip in the brain of a 25-year-old man known as MN, who had been left paralysed from the neck down by a knife wound.
Over the subsequent nine months, MN successfully used this BrainGate chip to open emails, operate a television and even control a robotic arm (Nature, vol 442, p 164). It was a promising step, but the technology was far from perfect. "Although BrainGate1 worked well in many ways, at times the control was not satisfactory," says Donoghue. And by the end of the trial, fluids from the brain had degraded the chip. The team are now solving these problems, and earlier this year announced the start of a clinical trial for an improved version of the chip.
With a chip implanted in his brain, a paralysed man was able to open emails, operate the TV and even control a robotic arm
The ultimate hope for many paralysed people, of course, is to regain movement in their own limbs. Until Fetz's experiment last year, no one had successfully used an implant to bridge a broken connection between the brain and the body. Trials of functional electrical stimulation (FES), in which implanted electrodes directly stimulate muscles into action, had hinted that this might be possible. But these impulses had been activated by external triggers, such as a switch controlled by one of the patient's healthy limbs, and not directly by brain signals.
Not only did Fetz's work demonstrate that the electronics could descramble neural signals and relay appropriate instructions to the limbs using FES, he also showed that the brain makes the job easier than one might expect. Although the motor neurons that connected to the chip did not naturally control the wrist, in a short time they adapted to the task and controlled complex actions (Nature, vol 456, p 639). "All neurons could be used equally well for control regardless of their original association to movement," says team member Chet Moritz.
That could have an important implication for humans hoping to use similar implants in the future. "It underscores the impressive flexibility of the brain in learning to adapt to novel connections, which may play a key role in allowing neural prostheses to be adopted by patients," he says.
So could the same approach work in humans? There seem to be no fundamental obstacles, and Donoghue plans to test the proposition in the new BrainGate trials, using his chip to control a limb using FES. If successful, it will represent a milestone in the development of such treatments.
Direct electrical stimulation of muscles using FES is unlikely to be the final solution, however. This direct approach uses a relatively powerful electric current applied to large areas of tissue, producing fairly clumsy movements. A more elegant method, some claim, is to send the impulse along the existing healthy nerves. That would require smaller local currents, delivered with greater precision, to finer regions of the muscle tissue, which should allow more subtle control.
As a bonus, nerve stimulation could simplify some of the demands placed on a brain chip. That's because for many rhythmic activities, such as breathing, walking and crawling, the brain simply sends a command signal and it is the spinal cord's in-built systems that orchestrate the fine movements of each muscle. So if the healthy sections of a damaged spinal cord have retained their ability to control movement, the electronic chip could transmit the brain signal around the broken connection but leave the muscular orchestration to the spinal cord. In this case, a brain chip would just beam the message to a second device implanted in the spine below the break, which would then stimulate the spinal cord.
The chips could simply transmit the information around the break, leaving the undamaged sections of the spinal cord to orchestrate the muscles
That could "dramatically simplify the control signals needed from the brain", says Moritz, since for these repetitive tasks the brain chip would just decode and transmit an umbrella command. Such simplification should make the chips less likely to fail - an important consideration when the only way to replace the chips is through invasive surgery - and also reduce their power consumption.
Using this principle in 2002, Vivian Mushahwar, now at the University of Alberta in Edmonton, Canada, plugged four electrodes into a cat's spinal cord and delivered signals that mimicked the brain's command to walk. Sure enough, the cat made stepping motions.
Simply relaying the messages across a break in this way would not help the worst injuries, however, in which the spinal cord has lost its ability to coordinate muscles. In these cases, to minimise the size of the brain chip, and the burden placed on it, the muscular orchestration would need to come from either the chip implanted in the spinal cord, or an external device that communicates wirelessly with the chips in the brain and the spine.
Calculating exactly which nerves to stimulate and in what pattern is no easy task, but the first demonstration of an artificial "central pattern generator" was reported last year, when Mushahwar and colleagues at Johns Hopkins University in Baltimore, Maryland, successfully tested such a chip on a cat. With coordination coming solely from an external CPG chip connected to a handful of electrodes that stimulated the cat's spine, the animal was able to walk (IEEE Transactions on Biomedical Circuits and Systems, vol 2, p 212). In this experiment, the team were simply testing the CPG's ability to orchestrate movement as an alternative to FES, so the trigger came from a manual switch and not the cat's brain. The next hurdle will be to use the CPG in conjunction with a neural chip.
While this CPG chip only dealt with the action of walking, in humans an additional external chip might also offload some of the processing from the brain chip for non-repetitive motions like clenching a fist or raising a hand. The brain doesn't necessarily produce an umbrella command for all of these movements, so the neural implant would still need to detect a more complicated signal, but the external chip could at least perform some of the processing to decode and relay these comands to the relevant electrodes.
For many patients, technology like this would only solve half the problem, however. Paralysed people who have lost feeling as well as movement in their limbs would need two-way systems to pass sensations back to their brain. This information could come from artificial sensors, but ideally the chip would read sensations from existing nerves and relay them to chips that stimulate the areas of the brain that process tactile information.
Although work has been slower in this area, there's good evidence it will one day be possible. Carmena, for instance, who is now at the University of California, Berkeley, recently stimulated a rat's brain to feel sensations from some "virtual whiskers", causing it to move as if its own whisker's had really brushed against an object. Similar technology could one day relay tactile information to human brains.
If these advances in brain-chip capability are to be exploited, the researchers still need to ensure that the chips are safe and durable. Biocompatibility, for instance, is a huge challenge, because tissue in the brain can react badly to an implant, killing off the very neurons that the electronics are trying to connect to. Recent efforts suggest a coating of growth hormones might mitigate this problem, while others have shown chips that slowly exude stem cells might also work.
Then there's the problem of powering the devices. Most existing implants - like cochlear implants, for example - are connected to a battery outside the head that can be replaced regularly. The electrodes in the spine and limbs could be powered this way, but it's less practical for a chip deep within the skull. Instead, such chips will need to be recharged by electromagnetic fields generated by a device outside the head, so power consumption will have to be minimal.
One solution might be to offload the more difficult processing to a portable computer outside the body, before passing the information back to the chips that stimulate the nervous system. In this way, Reid Harrison at the University of Utah in Salt Lake City has produced a neural chip that uses just 8 milliwatts. That's less than the "standby" LED on the front of a TV set.
All the pieces are gradually coming together, but whatever happens it will be a long time before these chips can become a mainstream treatment: the US Food and Drug Administration requires as much as 10 years of animal testing before a chip can be deemed safe enough to be implanted in human brains. That means the latest technology, such as chips that stimulate tactile sensations in the brain, will need extensive testing before clinical trials can begin.
Yet even once the technology has proven itself, the social issues surrounding the treatment will need to be solved. Take the question of security, for example. Last year, a team of researchers successfully hacked into a heart pacemaker and defibrillator through the wireless communication that allows doctors to adjust its performance. Although the device wasn't implanted in anyone at the time, it raised the possibility that hackers could disrupt a patient's treatment (New Scientist, 22 March 2008, p 23).
To make matters worse, there is currently no obvious way of protecting a defibrillator or pacemaker from a hacker without inhibiting a doctor from accessing it during an emergency. Since neural prostheses will rely so heavily on wireless links to communicate between the different components, the risk to these chips may be even greater.
Perhaps most perplexing is the question of legal responsibility. If someone wearing a neural prosthesis were to punch someone, who is to blame? The action may have been deliberate, in which case the patient is to blame, or the chip may have been malfunctioning and the responsibility would lie with the manufacturer. Discovering where the truth lay would be no easy task. The law has had trouble catching up with the self-parking car, never mind an electronically controlled limb gone wild.