Matt Nagle's extraordinary abilities were first reported in March 2005. Now details of the technology that lets him perform these tasks are published in the journal Nature. Another study in the same issue reveals a technique that could dramatically improve the speed with which such implants work.
Electrodes implanted in Nagle's brain measure the neural signals generated when he concentrates on trying to move one of his paralysed limbs. Software trained to recognise different patterns of neural activity then translates imagined gestures into the movement of an on-screen cursor or a robotic arm at Nagle's side.
"The fundamental findings are that you can record activity from the brain years after injury, that thinking about movement is sufficient to activate the brain, and that we can decode the signal," says John Donoghue of Brown University in Providence, Rhode Island, who led the work.
"Even though only one person was studied, the findings are impressive, especially as you can use the system while talking," says Maria Stokes a neurologist at the University of Southampton, UK.
Surgeons drilled a hole in Nagle's skull and inserted a pill-sized chip covered with 96 protruding electrodes into his motor cortex, an area at the centre of his brain that normally controls bodily movement. The operation carries the risk of infection and brain damage, an especially chilling prospect for someone already quadriplegic.
Billions of neurons
The motor cortex is made up of billions of neurons that constantly produce electrical signals, known as "spikes". Activity in different parts of the motor cortex was already known to correspond to different limbs, but Donoghue's team was able to identify signals related to different movements of the same limb.
To do this, they asked Nagle to imagine moving a limb and recorded the corresponding electrical signals. They found that the same small clusters of neurons produced different numbers of spikes when imagined moving a limb in different ways. For example, the same cluster might produce 15 spikes when he thought about moving the left arm to the left and only six when he thought about moving it to the right.
Donoghue's team built a list of spike patterns and corresponding movements. By automatically matching these together, a connected computer allows Nagle to control his computer or robot arm. The implant and computer were developed by a company set up by Donoghue himself, called Cyberkinetics, based in Massachusetts, US.
In the same issue of Nature, Krishna Shenoy and colleagues at Stanford University in California, US, report a way to dramatically boost the efficiency of brain implants in monkeys. Using software that predicts the monkey's intention from only the first few bursts of neural activity, the animals' implants were able to function four times faster than normal - a rate that could enable a paralysed person to type out 15 words per minute.
If the same techniques work in people, Donoghue's ambitious approach may yet be perfected. Working with researchers at Case Western Reserve University in Cleveland, Ohio, US, he ultimately hopes to enable quadriplegics to control their own limbs via electrical stimulation of their muscles.
Nagle's implant can only sample a fraction of the relevant brain activity and for the moment the signal can be patchy. This causes the cursor or robotic arm Nagle is controlling to wobble, and can make even simple tasks like checking email frustrating.
Less invasive brain-computer interfaces exist, but offer less control. A group led by Benjamin Blankertz at the Fraunhofer Institute in Berlin, Germany, has developed a system that measures electrical activity via a skull cap lined with electrodes. While this lets users type at a speed of eight characters a minute, it cannot perform more complex tasks like manipulating a robotic arm.
Journal reference: (Nature: vol 442, p 164; p 195).