杏吧原创

USING nothing but the power of thought, monkeys have been able to control an object in three dimensions as smoothly as if they were moving it with their own hands. The breakthrough, which could lead to artificial limbs controlled directly by the brain, reveals that visual feedback plays a crucial role in training the brain to do this.

Andrew Schwartz and his colleagues at Arizona State University in Tempe and the Neurosciences Institute in San Diego, California, started off by training rhesus macaques to use their hands to track a moving target. The monkeys were wearing a virtual reality headset, so they couldn鈥檛 see their hands. Instead, the position of their hands was represented by a coloured cursor that followed the target.

The monkeys鈥� arms were then tied down so that they couldn鈥檛 move. Instead, the researchers used signals from electrodes implanted in the monkeys鈥� brains to control the cursor as the monkeys tried to make it track the moving targets. The electrodes tapped into the motor neurons that are normally responsible for controlling the monkeys鈥� arm movement. Signals from just a few dozen electrodes were enough to track the movements, the team discovered.

What鈥檚 more, the monkeys鈥� accuracy improved as time went on. 鈥淵ou can see the animal homing in on the target,鈥� says Schwartz. 鈥淎fter about 20 to 30 days the monkey is moving the cursor almost as fast as its own hand would move. That鈥檚 our goal, to make movements like human movements.鈥�

It wasn鈥檛 only the accuracy that changed. 鈥淭he other amazing thing is that they observed that the behaviours of these neurons were also changing,鈥� says Paul Verschure, who works on developing neurological interfaces for machines at the Institute of Neuroinformatics in Zurich, Switzerland. This suggests that the brain had figured out it didn鈥檛 have to use all the neurons required to actually move the arm in order to move the cursor. Instead it had tuned in to use only those neurons that were necessary. Indeed, Schwartz noticed that after a while one of the monkey鈥檚 arms stopped moving every time the cursor moved. Verschure adds that this retuning of neurons could mean that different parts of the brain could be trained to produce useful signals for artificial limbs. This would be particularly useful for people with damage to the brain regions that normally control arm movement.

Previously, researchers have shown that signals from monkey neurons could be used to move a robotic arm which the monkey could not see, roughly in synchrony with the monkey鈥檚 own arm. In the latest experiment, visual feedback, like seeing the cursor move as they think, allows the monkey to fine-tune its movements, says Verschure.

It was previously assumed that signals from at least 600 implanted electrodes would be needed to get sufficient accuracy for practical applications. But giving the brain visual feedback seems to drastically cut the number needed. Though the idea of providing visual feedback seems obvious, many people didn鈥檛 look at the brain as part of a loop with the real world, says Verschure.

The implications for this research is that 鈥渘europrosthetic鈥� limbs controlled directly by the brain could become a reality, using today鈥檚 technology, says Verschure. 鈥淭hese guys have shown for the first time that this is a workable option,鈥� he says. 鈥淚t doesn鈥檛 need a big technological leap to turn it into a practical application.鈥� The results mean that there is no need for sophisticated electrodes and training algorithms to translate neural activity into signals that could usefully move artificial limbs, says Schwartz.

But there are still some obstacles to overcome, he says. Until long-term studies show that large numbers of electrodes permanently implanted into the brains of humans are both safe and reliable, neuroprosthetics will not be an option. And while the electrodes used by Schwartz appear to be sensitive enough for the job in hand, they frequently don鈥檛 work once implanted because they become damaged or are simply too far away from the neurons to pick up a good signal.