Of cells and wires
The first step to computer augmentation and neuroprosthetics lies in the connection between nerve cell and metal. How are scientists bridging the gap?
By Edyta Zielinska
Illustrations by Thom Graves
Neural probe illustrations by Christopher Burke. Provided courtesy of NeuroNexus Technologies, Ann Arbor, Michigan.
he man skis down sharp inclines at tremendous speeds, sees wind frolic through a woman's hair as the French countryside passes outside of the car window, checks out a nurse's cleavage. These are the visions and memories of the protagonist in The Diving Bell and the Butterfly, a film by Julian Shnabel about a man with locked-in-syndrome whose vibrant mind can only control the movement of his left eye. The audience experiences locked-in syndrome through the thoughts of the witty and irreverent Jean-Dominique Bauby, the former journalist and editor of the fashion magazine Elle, as he learns to communicate, and in fact dictate a bestselling memoir, with only the blink of his eye. Bauby became "locked-in" in 1995 and died in 1997 from pneumonia, seven years before the first man with locked-in syndrome was implanted with an electrode that might one day allow him to control a voice synthesizer with his thoughts.
1 Many in the field think the lack of direct contact between the probe and neurons is the major reason for the loss in continual recording.
Biomedical engineers are trying to bridge this 100-µm gap using a number of different approaches. Some groups have covered the electrodes with laminins, factors that are used to culture and attract neurons. Philip Kennedy's team at Neural Signals, Inc., which implanted the first electrode into the brain of a locked-in syndrome patient, modifies the technique by encasing recording wires with a glass tube, which protect the wires from attack by the immune system. Neurons can then sneak into the tube and interact directly with the wires.
Otto's research focuses on finding and coating the electrode with molecules that "convince the glia [and other reactive cells] to leave us alone." Other groups are working to make electrodes out of a softer material that might not disrupt as much of the brain cells upon entry. The problem, says Otto, is that the electrode must still be stiff enough to penetrate the brain, "which is not as much like Jell-O or soup as you think it is."
The standard electrodes inserted in the brain are the Utah and Michigan electrodes. With around 100 prongs, the Utah electrode, one of the "granddaddies" of neuro-electrodes, is rather large; when sitting atop a penny, it roughly covers Lincoln's brain.
A number of companies are developing the next generation of electrodes that can send signals wirelessly, which could reduce the tissue damage caused by the fork in the Jell-O problem, since electrodes would not be tethered to a fixed spot. "We'd like to make a fork with no handle that can move around with the Jell-O," says Reid Harrison at the University of Utah, who is designing a wireless device he says could be ready for clinical application in a decade. Engineers are also exploring different shapes and materials for the electrode.
David Martin's group at the University of Michigan has been developing an electrode in which a liquid seeps out of the tip and then polymerizes throughout the tissue surrounding the electrode, recording from the neurons that have retreated from direct contact with the electrode (see infographic below).2 Once set, the polymer and the surrounding neurons are "are so intimately interfaced that you can't pull them apart," says Sarah Richardson-Burns, a former postdoc of Martin's and currently the director of research and development at Biotectix, the company that is commercializing the technology. The company continues to test the electrode to see if it can continuously record, she adds.
The Utah electrode, designed in 1989 by Richard Normann at the University of Utah, has a good track record. It's almost identical to the electrode Donoghue's group used to continuously record from human patients for over three years ("We had a cake at 1,000 days," says Donoghue). "There's no hard quantitative explanation for [the success or failure of] any one case," says Daryl Kipke from the University of Michigan, who works on improving the reliability of implanted neural electrodes. "Every good implantable electrode has had at least some example of lasting months or years," says Kipke. Donoghue says one reason the signal from his electrodes lasted so long was the fact that the electrode platform sits atop the brain—its 100 prongs penetrate the tissue, and only slack wire leads out of the brain. There's no hard "fork handle" component fixed to the skull, so the electrode "bounces with your brain," he says. In his experience, the biggest problems with electrodes failing have been technical, not biological. "We are quite pleased that tissue reaction [to the electrode] is not the formidable problem we thought it was."
Meanwhile, some scientists are using a less invasive approach to monitor neural activity and thereby circumvent the problems of implanted electrodes. Gerwin Schalk at the Wadsworth Center in New York City has been working on adapting a technique used by some hospitals to detect where seizures originate in the brain. Doctors place electrodes under the skull atop the surface of the brain to scan for seizure activity. In Schalk's version, most of the electrodes will be placed over the motor cortex. Because the technique is less invasive, the signal usually can be picked up for longer periods of time. However, because the technique records from many neurons at a distance, the signal quality is considerably lower than implanted electrodes, says Kipke.
For more than 20 years, researchers at the Cleveland FES Center have been training patients with paralysis to control their own neurological signals—not from electrodes in their brains, but from sensors implanted in muscles that they can still control. When the patient moves those muscles, the implanted sensor picks up the signal and translates it using a brick-sized computer. The computer-translated signal is routed to electrical stimulators implanted at the nerves or directly on the paralyzed muscles, causing them to move. In other words, by twitching muscles in her neck, a patient can lift his or her paralyzed arm. By twitching another muscle, the patient can move that arm towards the left.3 (Watch the video online at the-scientist.com.)
Patients can even achieve such complex actions as grasping a cup, says Hunter Peckham, director of the FES Center. Rather than thinking about contracting each individual hand muscle necessary for grasping, a patient activates one "grasp" muscle, which sends a signal to the computer to fire stimulation into all 12-16 muscles needed for closing the hand.
These systems have been implanted in over 300 patients to restore functions such as bladder control, grasping, standing, walking and even penile erection, all part of the center's research program. But each acts as an individual modality; depending on the level of paralysis, a patient has to choose the function he or she wants to restore most, since there may not be many non-paralyzed muscles available for sensor implantation. "We're sort of agnostic about where the [control] signal comes from," says Peckham, indicating that he would consider sources besides non-paralyzed muscles. Going into the brain for control signals rather than muscles could increase the number of sources and theoretically allow a patient to control many functions simultaneously.
To investigate that possibility, FES's Kirsch recently teamed up with John Donoghue's group at Brown University. Donoghue's team has implanted electrodes into the brains of four patients who then used their thoughts—as translated by an external computer—to move a cursor on a computer screen or control a robotic arm in a simple open/close action.
The first goal of the collaboration is to see how well a patient can use her mind to control an arm she sees on a computer screen via the electrode implanted in her brain. Only after a patient demonstrates an ability to control the virtual arm with her thoughts—and a wireless brain electrode becomes available—will Kirsch's team consider implanting electrical stimulators into the arms of the patients to control her own arm (see infographic on this page). A recent patient of Donoghue's demonstrated the ability to control a wheelchair with her thoughts, though it was not the one she was sitting in—a dry run for future experiments in which she could hopefully control her own chair. Researchers have already shown that this type of technique can work in monkeys, and in a much simpler system than Kirsch and Donoghue's. Instead of using a multi-pronged electrode which records from many neurons, Chet Moritz and colleagues from the University of Washington recorded signals from one neuron at a time in the brain of a monkey. The monkey quickly learned to perform a simple forward/backward motion in a paralyzed hand, learning the controls within 10 to 30 minutes. The monkey also figured out how to use its brain to control the force of the movement.
When recording from single neurons, the problem will be maintaining consistency between days, says Kirsch. "If I'm surveying all Democrats," the consensus opinion will be fairly unified, and "the opinion from day to day won't change that much," Kirsch says. But on any given day, he adds, the opinion of one Democrat could waver quite a bit. In the same way, recording from single neurons could produce a signal that varies more than the signal coming from a group of neurons. What finally goes into humans "will probably be some combination of the two" signal translations, predicts Leigh Hochberg, a clinician and researcher at Brown University who works with Donoghue.
The future of brain-machine interfaces will have to be fully implantable, similar to the cochlear implant, which has already helped the hearing-impaired by translating sound picked up by a speaker into electrical stimulation of the auditory nerve within the cochlea. The problem is power and size. The reason the cochlear implant works so well is that the signal is fairly simple and it stimulates the brain, without recording from neurons. "Stimulation requires less eloquence than recording," says Otto.
With implanted (ultimately wireless) brain electrodes, "the amount of data" that will need to be transmitted "can be gigantic," says Kirsch, increasing the need for power. A coil that receives power from a battery pack outside the brain, plus a processor to translate the signals, "don't fit as nicely in that space between your brain and your skull," says Kirsch. Another problem is that any data processor tends to heat up, a potentially killing off nearby neurons in the brain, says Harrison. He has designed a microprocessor to fit on the back of a Utah electrode, which essentially digests the recorded information from the electrodes into only the "spikes" of the neuronal action potential. These neuron-digests decrease the need for power while still predicting motor activity quite well.
But brain recording coupled with stimulation—that is, also sending signals back to the brain—is the future of neuroprosthetics. Such an advance would help humans sense or perceive their prosthetic motions without looking. Harrison is working on implanting stimulating electrodes—a modified Utah electrode—to help amputees "feel" their prosthetic hand, a technology that might be ready for use in humans in a decade. It would be useful to know how "softly they're touching something," says Harrison. But the process will be one of trial and error. It's not hard to implant an electrode into a nerve bundle, "but we don't know which fiber" in that bundle will give the sensation of touch or weight or heat, since that varies for every individual, says Harrison. The stimulation "would probably feel weird," says Otto, but it would be enough to sense what a limb is doing without looking at it.
Others are trying to send signals back to the brain by creating something akin to prosthetic neurons, which restore lost cognitive function to damaged brain tissue.4 One group led by Theodore Berger at the University of Southern California is developing methods to restore connections in the hippocampal region of the brain, where long term memories such as those that allow people to make associations between names and faces are stored, and which houses the neurons that degenerate in Alzheimer's disease. "Anything involving memory is going to be in the far, far future," says Harrison. But, as they say, the future gets closer every day. "The sky is the limit with brain-machine interfaces," says Otto.