Cybermonkeys' robotic arms bode well for artificial limbs

By Jonathan M. Gitlin | Published: May 28, 2008 - 12:00PM CT

Three years ago, we carried a brief report about a group of researchers in Pittsburgh who were working on a robotic arm that was being neurally controlled by a monkey. That work is published today in Nature, and it has been worth the wait. I'll leave the comedy potential of cybermonkeys until the end and go into a bit of the science first.

Related Stories

• Artifical hands that feel

The neural control of movement has many components, but the motor cortex is central to them. The motor cortex has several areas, but the primary motor cortex, a strip of brain just behind the frontal lobe, is where you decide to send out signals to peripheral nerves to tell your arms or legs to do something. The cortex is also organized somatotopically: the region that controls the feet is next to that which controls the leg, then trunk, arm, and so on. The practical upshot of this is that we have a good idea which region of the primary motor cortex needs to be measured in order to work on controlling a cybernetic arm or hand.

Microelectrodes were implanted in the regions that control part of the monkeys' arm, and they were trained to use neural activity to control the prosthetic arm. Since you can't really give a monkey instructions, this was achieved by training the monkeys with a one-dimensional joystick; as they pushed the joystick, the arm would move. After several days, the monkeys had enough control with the joystick to retrieve food and feed themselves. Over time, the input from the joystick was reduced and cortical control was developed.

Since the arm has more degrees of freedom that need to be controlled compared to the one-dimensional joystick, the training was accomplished in stages. One priority for the researchers was to ensure that the delay between activation of the cortical region and movement of the arm wasn't excessive. Detecting the signal, processing it, and having the arm move took around 150 milliseconds, which isn't too far off the delay between activation of the motor cortex and moving an actual limb.

Although the monkeys were only trained in four tasks (move the arm to the food, grasp the food, move the arm to the mouth, ungrasp the food), monkey A (which was controlling a more advanced arm) refined its method, beginning to ungrasp sticky food after retrieving it, presumably after recognizing that it needn't be held unnecessarily. Monkey A was able to self-feed successfully 61 percent of the time; not a bad success rate for a complex procedure.

Previous experiments centered on using the primary motor cortex as a brain/computer interface have used cortical activity to control a cursor on a screen. This study instead uses brain activity to control a physical object that the monkey can interact with. The authors argue that this added factor is a major element of their success, and that it bodes well for the development of neurally-controlled artificial limbs. Cortical control, combined with tactile feedback similar to that described last year, should make for functional cybernetic limbs with a minimum of wires protruding from one's scalp.

Okay, commence cybermonkey jokes now.