Producing accurate and stable, long-term readings of neuronal activity using a brain–machine interface (BMI) is now possible thanks to work by Naotaka Fujii and his colleagues at the RIKEN Brain Science Institute, Wako1. Their results could help researchers to develop durable and versatile neural prostheses for rehabilitation patients.
BMIs read neural activity associated with planning and executing movements and decode it into commands that are relayed to an external device such as a computer cursor or robotic arm. This normally involves recording simultaneously from multiple, single neurons, so the recordings are unstable and the decoding model needs re-calibration on a daily basis.
Fujii and colleagues used an alternative technique called electrocorticography, in which an array of electrodes is used to record the population activity of cortical neurons.
Electrocorticography is often used to evaluate epileptic patients before neurosurgery but is not normally used for longer than two weeks. It was thought to provide a low fidelity signal for BMIs, because the electrodes record neural activity from the cortical surface, rather than within the cortex.
To overcome this, the researchers designed an electrode array for long-term recording, and developed a novel decoding algorithm that samples neural activity from multiple brain regions.
After implanting the electrodes into the brains of monkeys, so that they spanned multiple brain regions, Fujii and colleagues trained the animals to spontaneously reach out and grasp food presented to them. The monkeys wore custom-made jackets fitted with reflective markers at the shoulders, elbows and wrists. The researchers then recorded the monkeys’ arm movements using a motion capture system, and correlated them with the neuronal activity recorded by the electrodes.
By decoding the signals, they could predict the trajectory and orientation of the monkeys’ arms in three dimensions. The accuracy of the decoding was comparable to that of existing BMIs which record activity from single cells. Furthermore, the recordings were highly stable, and could be decoded for several months without recalibration.
The new recording technique should prove to be useful for researchers investigating movement control and higher cognitive functions. It could also lead to versatile devices that can be implanted for long periods of time, to aid patients with brain damage, spinal cord injury, and neurodegenerative conditions such as amyotrophic lateral sclerosis, notes Fujii.
“Our electrode array is still not ready for long-term use in patients, because of the risk of infection,” says Fujii, “but we are now developing a fully implantable wireless device to prevent this.”
The corresponding author for this highlight is based at the Laboratory for Adaptive Intelligence, RIKEN Brain Science Institute
