A robotic hand controlled by the power of thought alone has been demonstrated by researchers in Japan.

The robotic hand mimics the movements of a person's real hand, based on real-time functional magnetic resonance imaging (fMRI) of their brain activity. It marks another landmark in the advance towards prosthetics and computers that can be operating by thought alone.

The system was developed by Yukiyasu Kamitani and colleagues from the ATR Computational Neuroscience Laboratories in Kyoto, and researchers from the Honda Research Institute in Saitama.

Subjects lay inside an MRI scanner and were asked to make "rock, paper, scissor" shapes with their right hand. As they did this, the MRI scanner recorded brain activity during the formation of each shape and fed this data to a connected computer. After a short training period, the computer was able to recognise the brain activity associated with each shape and command the robotic appendage do the same.

Magnetic field

An fMRI machine probes activity within the brain by monitoring blood flow to different regions. It uses a powerful magnetic field combined with radiofrequency pulses to probe the magnetic state of hydrogen atoms in water molecules within body tissue.

An alternative and more portable method is to measure electrical activity inside the brain using electrodes either implanted in brain tissue or attached to the scalp. US researchers have previously used brain implants to allow monkeys to remotely operate robotic arms.

Electrodes attached to a person's skull can also be used to control the movement of a cursor across a computer screen. Klaus-Robert Mueller at the Fraunhofer Institute in Berlin, Germany, has developed such a system. He says the fMRI technique is cumbersome and expensive but could help scientists better understand how the brain works because it provides higher resolution.

Reading intent

"From a practical point of view the technology is too costly and slow," he told New Scientist. "But it's very interesting that you can do something as complicated as this."

Kamitani and colleagues have previously shown that fMRI scanning can be used to distinguish between simple images that a subject is looking at and thinking about.

One day, Kamitani believes, the robot hand could be made to respond faster than a user's real one. "The next step for me is to decode faster, even before the person moves their hand, by reading the brain activity related to intention," he told New Scientist.

But he admits that fMRI scanning technology must be improved dramatically before this could be possible, and before the system could be used practically. "We will need several breakthroughs in related technologies, including those for brain scanning hardware, before this type of non-invasive systems will be used in daily life," he says.