With the development of a new type of brain–machine interface users in the future could be able to plan and perform a series of sequential movements more naturally. “This is different from BMIs used thus far, which require users to plan and execute each element of the sequential movement one at a time. In such BMIs, for example, the user cannot plan the second letter of a word before typing the first letter,”, said Ziv Williams, co-author of the study, referring probably to today’s brain spellers such as the Intendix.
“Development of this new BMI implies that it may be possible, in principle, for patients to plan and perform sequential movements as they would do naturally, for example typing the full planned series of letters in a word”.
Williams said the new design could also lead to the development of BMIs that can analyse intended movements before executing them – therefore enabling a robotic limb to perform the movements more effectively.
“For example, it may also be able to type the word faster or remove spelling mistakes before typing,” he said.
Williams added that the next step would be to further examine how the new BMI can be used to perform more fluid and accurate sequential movements using a robotic limb, and to test the design on human patients.
With 50–80% accuracy there is still plenty to do
During the study Williams and his colleagues recorded the electrical impulses from rhesus monkeys’ brains trained to remember a sequence of two locations on a computer screen and, after a short pause, move the cursor to those locations.
The scientists found that the two movements could be decoded, using computer algorithms, from separate, small groups of neurones in the premotor cortex – a part of the brain involved in planning and executing limb movements.
The two distinct groups of neurones allowed the two planned targets of the movement to be simultaneously held, without degradation, in the ‘working’ memory – a brain system that provides temporary storage and real-time processing of the information necessary to perform complex tasks.
Exploiting these mechanisms, the team then developed a BMI that could not only predict both of the intended movements simultaneously, but also drive the movements in real time alongside the monkey’s motor response.
However, there is still plenty more work to be done. According to Geoffrey Goodhill, a computational neuroscientist at The University of Queensland, Australia, “The decoder only correctly predicted the intended movement 50–80% of the time. This was much better than chance, but certainly not good enough to, for instance, type fluently on a keyboard,” he said.