Brain-muscle interface helps paralysed monkeys move

Researchers from the University of Washington have demonstrated that paralysed monkeys can move using a simple neuroprosthesis consisting of an external electrical circuit which connects individual neurons in the motor cortex to muscles in the arm.

Similar prostheses have been used to move external devices such as a robotic arm, but they required sophisticated algorithms to decode the brain activity associated with generating movements. The researchers adopted the alternative strategy of creating an artificial connection by which the activity of single cells could directly stimulate the muscles.

The new study, which is published online in the journal Nature, is a significant advance, as it provides proof of principle that a similar "brain-muscle interface" could eventually be used as a treatment for patients paralysed by spinal cord injuries.

The prosthesis developed by Eberhard Fetz and his colleagues at the Washington National Primate Research Center in Seattle consists of  an array of 12 tungsten microwire recording electrodes, each of which can be moved independently of the others by means of tiny piezoelectric motors. Previous work has shown that this electrode array can be used to obtain stable recordings from the same neurons for around 1 week and, by periodically moving individual electrodes, to systematically sample large numbers of new cells for periods of 1 year or more.


(From Jackson & Fetz, 2007)

In the new study, the device was implanted into the arm and wrist region of the primary motor cortex of two macaque monkeys (which were located using the technique developed in the 1930s by Wilder Penfield). The animals were trained to move a cursor towards targets on a computer screen, using rotational movements (or torque) of the wrist, in return for a reward of apple sauce. During this training phase, the researchers identified some cells which control flexor muscles and others which control extensors. Some of the cells were also found to be associated with rotation in a specific direction.

Next, the researchers temporarily paralysed the animals' wrists by injecting a local anaesthetic into catheters inserted into the median, ulnar and radial nerves which run down the arm. This effectively blocked all nerve activity, so that the signals generated in the spinal cord could not reach the arm muscles which generate wrist torque. As a result, the monkeys were unable to perform the movements required to execute the computer task.

However, when the activity recorded by the electrodes was re-routed past the anaesthetized nerves and directly to 8 muscles in the arm, movement was restored. A burst of activity from a single neuron was found to be sufficient to stimulate muscle contraction and generate enough torque for the monkeys to move the cursor. Furthermore, the monkeys learnt very quickly how to use the device, and were able to use it to control multiple muscles simultaneously. Remarkably, this was the case for almost all of the motor neurons from which recordings were taken, regardless of whether or not they associated with wrist movement. Activity from the same cell could stimulate antagonist pairs of muscles, to generate torque which made the wrist rotate first one way and then the other.

This is a significant advance over similar prosthetic devices which have recently been used by monkeys and quadriplegic patients to control the movements of robotic arms. These earlier devices record the combined activity of populations of nerve cells in the premotor cortex, which is involved in planning movements, and send it to a computer, which decodes the patterns associated with a particular task and translates them into executable commands. By contrast, this device records from single cells in the primary motor cortex which are not necessarily associated with the required movements. It therefore does not rely on complex decoding algorithms, but instead has a neurochip which converts the activity into electrical impulses that stimulate the muscle.  

With other chronically implanted electrode arrays, the tissue response decreases with time. This device allows for stable recordings over long periods of time, possibly because the electrodes can be placed accurately near the large pyramidal cells in layer 5 of the cerebral cortex, whose axons form a bundle of fibres that descends to the motor neurons in the spinal cord which control the muscles. Also, the ability to repeatedly adjust the depth of each electrode by small incremental movements, so that it records from progressively deeper cells, likely reduces the damage to the tissue.

Chronic implantation is used routinely in small mammals such as rats, to investigate, for example, the role of place cells in spatial navigation. In larger mammals, such as monkeys or humans, it is more difficult to obtain good recordings for extended periods, because the membranes enveloping the brain are tougher and because the spaces beneath them are larger, which causes the brain to move more relative to the skull. There are bigger obstacles to the development of a neuroprosthesis for patients with spinal cord injury. For example, such a device would only be effective if it could generate complex sequential patterns of muscle contractions.

The electrode array developed by Fetz and his colleagues has many advantages over other brain-computer interfaces, but it has not been tested in permanently paralyzed animals. So although the new findings are notable, they will not lead to clinical applications any time soon. In the meantime, the device could prove to be very useful for investigating the long-term changes in neuronal activity which are associated with processes such as learning and memory.

Moritz, C. T., et al (2008). Direct control of paralysed muscles by cortical neurons. Nature. DOI: 10.1038/nature07418.

Jackson, A. & Fetz, E. E. (2007). Compact movable microwire array for long-term chronic unit recording in cerebral cortex of primates. J. Neurophysiol. 98: 3109-3118. DOI: 10.1152/jn.00569.2007.

More like this

Thanks for the update - that's a very clever experiment!

I doubt this particular electrode design would be useful for a clinical BMI. It certainly seems useful for research purposes, though the general idea is nothing new -- Current Protocols in Neuroscience has instructions on how to build a simple non-motorised version without any special equipment.

The Utah-bed-of-nails-style arrays used by Donoghue and others provide good recording stability as they have an integrated flexible ribbon cable, which allows the array to "float" in the brain, relative to movement of the skull-mounted connector. I think rather than movable arrays, we'll see higher density 3D Utah-style arrays, perhaps with coatings to limit tissue growth, or even neurotrophins to encourage neural processes to grow onto the electrodes.