Monkey See, Robot Do
Part of what limits EEG signals is that when one thinks about moving a cursor, or an arm, thousands of brain cells fire off simultaneously. Surface electrodes pick up all of the brain waves at once in a cacophony of electrical activity. That’s why a growing number of researchers are working on what’s termed “invasive” brain-computer interfaces. By tapping directly into the motor cortex, they hope that they can get past the EEG’s cocktail party chatter to tune into individual neurons, an advance they think could be key to helping the paralyzed operate FES devices.
Already, a number of animal experiments are suggesting this is precisely the case. In a startling result published last summer in the journal Nature Neuroscience, Duke’s Nicolelis and John Chapin, a neuroscientist at Hahnemann University in Philadelphia, reported that they had been able to get a rat to operate a robotic lever in real time via two dozen electrodes implanted in the area of the motor cortex that controls paw movement.
Several academic teams, including Chapin’s as well as groups at Brown University and the California Institute of Technology, are trying to reproduce similar results in monkeys, whose brains are more like our own. So far, some of the most exciting results have come from neurophysiologist Andrew Schwartz at San Diego’s Neurosciences Institute and collaborators at Arizona State University. Using dozens of hair-width electrodes implanted in the brain of a rhesus monkey, Schwartz simultaneously recorded signals from about 50 individual neurons, which he fed through a data-crunching algorithm to a robotic arm in a separate room. “And we see,” he says, “that the robotic arm moves close to the same trajectory that the monkey moved its arm.” A split-screen movie of the result can be seen on the Web at www.nsi.edu/motorlab.
This feat is possible even though scientists still know very little about how the brain creates movement. The trick, Schwartz explains, is that although there are millions of neurons in the motor cortex, measuring the “firing rates” of just a few cells can give a surprisingly accurate picture of where and how fast the monkey’s arm is moving. “It’s like doing a survey. You’re not going to get every person, but if you have enough samples you can get a pretty good idea of what’s going on,” he says.
Although Schwartz’s primates were unaware of the robot mimicking their movements, he’s now working on an experiment in which he’ll challenge restrained monkeys to use a thought-driven arm to feed themselves. A positive outcome would be proof-of-principle that a cortical signal could give quadriplegics precise control over FES devices like Freehand. In fact, Schwartz predicts that a rudimentary brain-activated robotic arm will be ready for human use within five years.
Even a successful human test won’t automatically translate into a working device. The development of invasive recording electrodes has been going on for some 30 years, but is still plagued with problems. In animal studies, signals from implanted electrodes tend to diminish over time, which may be due to scar tissue or shifting of the electrode caused by the brain’s normal movement within the skull. Schwartz calls “the long-term survival of the electrodes” a key problem, and admits that the Teflon-coated stainless steel wires he uses are “really crude devices.”
But improved electrodes is an engineering challenge that several teams are already looking to meet. Some of the most successful work to date has been accomplished by neurologist Phillip Kennedy of Atlanta, Ga., who was the first to implant cortical recording electrodes in a human being. And the Duke group has helped develop a matrix of 16 electrodes, just a square centimeter in area, which Plexon Inc., of Dallas, Texas, is now manufacturing. The electrodes are working well in primate experiments, but Nicolelis adds that “we need to evolve to a new generation.” Already looking ahead to applications in people, Duke is designing a telemetry chip to connect to the electrode array and transmit neuron recordings to an external computer, without wires coming through the skull.