Algorithm improves speed and accuracy of neural prostheses
Researchers in the US have improved the technology that allows paralysed people to control computers with their thoughts.
A team from Stanford University has developed an algorithm that is claimed to improve the speed and accuracy of the brain implants that pick up neural activity and translate it into cursor movement on a computer screen.
Most existing neural prostheses are based on brain activity data that was recorded as the subject moved or imagined moving his or her arm and was later analysed.
The new ReFIT algorithm was designed to incorporate information from the user (in this case, a monkey) as he or she was using it, including real-time feedback adjustments to the cursor’s movement (known as closed-loop decoding).
This allowed the scientists, led by Prof Krishna Shenoy, Dr Vikash Gilja and Paul Nuyujukian, to build the algorithm so it could learn from the user’s corrective movements, improving the speed and accuracy of the technology.
ReFIT can also interpret neural signals about both the cursor’s position and its velocity at once — something previous systems have not been able to do — further improving its movement.
Specifically, the new algorithm was better at stopping the cursor in the right place, as previous versions would often overshoot the target and it would take longer for the user to stop the cursor in the desired place.
To create such a responsive system, the team decided not to take the traditional approach of focusing on individual neurons.
‘From an engineering perspective, the process of isolating single neurons is difficult, due to minute physical movements between the electrode and nearby neurons, making it error prone,’ said Gilja.
This approach also increased the life of the device because neural implants that are fine-tuned to specific neurons degrade over time. This usually tends to happen within six months, but Gilja said the new device was still working after four years.
Commenting on the Stanford research, James Gnadt, a program director in systems and cognitive neuroscience at the National Institute of Neurological Disorders and Stroke, said: ‘Despite great progress in brain-computer interfaces to control the movement of devices such as prosthetic limbs, we’ve been left so far with halting, jerky, Etch-a-Sketch-like movements.
‘Shenoy’s study is a big step toward clinically useful brain-machine technology that has faster, smoother, more natural movements.’
Gilja added that the technology could eventually be developed to create robotic limbs that are controlled by the brain.