The first thing Ian Burkhart did after he got to the Outer Banks in North Carolina one afternoon in 2010 was dive into the ocean. Nineteen years old, Burkhart had just completed his freshman year of college and come to the beach to vacation with friends. Joyously and with abandon, he pitched himself into the waters.
As he swam and bobbed in the surf, a wave flung his body onto a sand bar, jerking his neck with a force stronger than anything he had experienced before. Burkhart found himself lying face down on the seafloor, struggling to move. “Stay calm,” he thought to himself. “Don't freak out.” The water was only a few feet deep where he lay, and his friends pulled him out. Laid flat on the sand, Burkhart still could not move. He lost consciousness as he was being carried into a helicopter that airlifted him to a trauma hospital in Virginia.
In a surgery that lasted almost nine hours, doctors put two rods into Burkhart's spine to stabilize it. The next day they delivered a grim diagnosis: the accident had broken two of the vertebrae at the base of his neck, damaging that section of his spinal cord. He would never be able to walk and would barely be able to move his arms again. He would need help with nearly all the things he had previously done with his brain on autopilot as a healthy 19-year-old: eating, going to the bathroom, holding a toothbrush, turning the knob of his car radio.
A wave of devastation crashed over Burkhart. Even as a child, he had strived to become self-reliant. He had played lacrosse since third grade and had been a Boy Scout. At 13, he took up delivering newspapers to make money. In high school, he had started a lawn-mowing company with his brother. Now, a year after entering college, he was looking at spending the rest of his life under round-the-clock care. That is when the magnitude of his loss began to hit him. “I thought, ‘Oh, crap, this is major,’” he says. “There was a lot of numbness because I didn't even know how to respond.”
But last year Burkhart did something neither he nor his doctors could have imagined him doing. Seated in a wheelchair inside a laboratory at the Ohio State University Wexner Medical Center in Columbus, he maneuvered his hand to pour from a bottle, stir his coffee with a straw and swipe a credit card through a reader, simply by thinking those actions. He achieved these feats using a chip implanted in his brain that transmitted the neural signals of his thoughts to a sleeve wrapped around his right forearm. Dotted with buttons that deliver tiny jolts of electricity to various muscles, the sleeve stimulated his hand to execute the movements he envisioned.
Burkhart's accomplishment represents a milestone in a decades-long effort to develop brain-computer interfaces to restore movement and other functions to individuals with spinal cord injuries. “These demonstrations are very impressive,” says Andrew Jackson, a neuroscientist at Newcastle University in England, who is not involved with the project. “This is a field where things are moving very quickly from monkey experiments to human experiments, and this was another study that showed that.”
The brain implant, the software interface between the chip and the computer, and the sleeve Burkhart wore on his arm are the culmination of decades of research in neuroscience, rehabilitation science, computer science and sensor design, as well as a preview of the technological breakthroughs that lie ahead. Although Burkhart was able to make those simple hand motions only within a lab, researchers hope the technology will one day enable spinal cord injury patients to regain the use of their limbs permanently, restoring their sense of normalcy and autonomy.
The advance, reported last year in Nature by researchers at Ohio State and the Battelle Memorial Institute, represents a triumph of both science and the human spirit. Burkhart has spent several hours each week over the past three years to help engineers perfect the algorithms that translate his brain signals into action. “Ian's the hero here,” says Ali Rezai, director of Ohio State's Neurological Institute and a member of the research team. “He has tremendous resilience and dedication. It's because of him that we are making these strides.”
Mind over Machine
After his surgery, Burkhart moved back home to Columbus and signed up for an outpatient rehabilitation program at Ohio State. He had always been optimistic by nature, and as he adjusted to life in a wheelchair, he concluded his only option was “to make the best of my situation.” His doctor was Jerry Mysiw, chair of the department of physical medicine and rehabilitation, who had spent more than 20 years working with spinal cord injury patients. Burkhart wanted to stay informed about research advances that could help patients like him. He had high hopes for “an advancement of some sort in my lifetime that would improve my daily life.”
Around the same time Burkhart was going to his rehab sessions, researchers at Ohio State were embarking on a collaboration with engineers at Battelle to translate neural signals from the brain into movements via a brain-computer interface. The idea of connecting the brain to a computer and converting the brain's electrical activity into actions had been around for more than half a century, but only in the past two decades have researchers landed on a workable approach, beginning with studies in lab animals.
In 1998 a chip developed by neuroscientist Philip Kennedy was implanted in a human patient for the first time, enabling the person to slowly move a cursor to spell out words on a computer screen. Starting in the late 1990s, Miguel Nicolelis and his colleagues at Duke University performed a series of experiments in which monkeys hooked up to an interface could, with training, control a robotic arm with their neural signals.
In the ensuing years, researchers were able to record signals from individual neurons or small groups of cells rather than a broad cacophony. This, together with improved machine-learning algorithms for interpreting the signals, made it possible to direct more intricate movements. A consortium of researchers led by neuroscientists John Donoghue and Leigh Hochberg of Brown University developed an interface called BrainGate that enables patients to move a cursor on a computer screen with their thoughts alone. In 2008 Andrew Schwartz of the University of Pittsburgh and his colleagues trained a monkey with an implant to mentally manipulate a robotic arm to feed itself marshmallows and fruit, demonstrating impressive dexterity. Both groups have since shown that humans with brain implants could control a robotic arm to perform similarly exacting feats.
But for patients, moving a robotic arm does not compare with the dream of regaining motion in one's own paralyzed limb. The Ohio State–Battelle team set its sights on achieving that dream by developing an arm-stimulating device that could communicate with a brain-computer interface. Mysiw asked Burkhart if he would participate in a study to test the sleeve, which Burkhart was glad to do. Every week at the lab beginning in September 2013, researchers slipped the stimulator onto his arm and hooked it up by wire to a computer. “We were able to get really good results as far as what type of movements we could make my hand do—such as flexing my fingers or making a fist—something that was very exciting and promising to me,” Burkhart says.
But these movements were the computer's, not his own. When the stimulator study was wrapping up, Mysiw sat down with Burkhart to explain the broader idea of the project: implanting a device in the brain that could control the stimulator directly. Would he like to volunteer for the procedure? Mysiw went over the implications with Burkhart. He would have to undergo at least two elective brain surgeries—one to insert the implant, another to take it out—with the inherent risk of infection. He would be exposing his brain to further damage, and even if the study were successful, he would not benefit from it personally, because the researchers would not be allowed to leave the implant in his brain permanently.
Burkhart felt the positives outweighed the negatives, however. He told Mysiw he felt it would be irresponsible to pass up such an opportunity to help others like him. In addition, Mysiw recalls, “he wanted to be able to scratch his nose [and] brush his teeth.” Even though Burkhart understood that the research would likely take a long time to yield a usable neural prosthesis, the prospect—however remote—was appealing. “Knowing there's a chance that I could use this in my everyday life, coupled with being able to make a big impact, really made me want to do it,” he says.
On April 22, 2014, Rezai and his fellow surgeons implanted a chip the size of a baby aspirin into Burkhart's motor cortex, in an area responsible for controlling his right hand. The chip is equipped with 96 tiny electrodes, each of which records the aggregate electrical activity of hundreds of nearby neurons. It is connected to a wire that sticks out of Burkhart's scalp, whose tip sits inside a nickel-sized disk screwed onto his crown. The structure resembles a small bottle cap and functions as the port through which researchers hook up Burkhart's brain to a computer in the lab. Learning to live with a metallic protrusion on his head took a while. “At first I had to really position my pillow just right so I wasn't putting any pressure on it,” he says. Over time, it came to feel less obtrusive.
Decoding the Language of Movement
In the years since undergoing the surgery, Burkhart has followed a demanding routine: spending several hours in the lab twice or three times a week to focus on making his hand do things it had lost the ability to do naturally. Every session starts the same way. A cable hooked up to his port relays the neural signals from his brain to a computer while researchers have him think about performing specific movements with his right hand—such as flexing a finger or clenching his fist. To focus on the tasks, he uses visual feedback from a virtual hand on a screen that is controlled by his brain signals.
The key challenge is to ensure that the computer correctly interprets the pattern of neural signals generated by Burkhart's thoughts. While millions of neurons are firing in his motor cortex every time he thinks of an action, the chip in his brain picks up and transmits only a tiny sample of those signals. At 30,000 samples per second for each of the 96 microelectrodes, the data are still voluminous. The first step is to compress this information without stripping it of meaning. The processed signals are then fed into a set of algorithms that filter and convert them into electrical commands for the muscles.
The decoder—as these algorithms are collectively called—compares the pattern of activity received from Burkhart's brain with previously recorded patterns corresponding to a variety of movements and resting states. By this process, the decoder determines what action Burkhart most likely intends to perform. The decoder then receives feedback on whether it judged the pattern correctly or not, helping it do better over successive trials.
The firing patterns corresponding to the same movement can look quite different from one session to another because of the natural day-to-day variability in brain activity, according to Gaurav Sharma, a research scientist at Battelle and the nonprofit's lead investigator for the project. Shifting of the implant inside the brain also contributes to variability. “Ian might be happy or sad, he might be tired, he might be hungry, he might be hot or cold,” Sharma says. And so, on any given day, the decoder needs to learn anew “to recognize the patterns that are consistent with Ian thinking, ‘I want to move my finger, I want to flex my wrist.’”
Burkhart went through a steep learning curve himself. When he began these sessions, a month after getting the implant, he did not know how he was supposed to mentally execute movements he had performed unconsciously before his accident. “The first few months of sessions, I would leave there being just completely mentally drained. I would feel like I [had] just sat through an all-day exam,” he says.
One day in June 2014, less than six weeks after the sessions had begun, Burkhart was able to clench and unclench his hand around a spoon just by thinking it. The researchers who were there to witness the moment whooped and cheered. “There was just so much excitement out of everybody in the room,” Burkhart says. Even before the cheers had died down, however, he was eager to hear what the next steps would be. “Now it's time to get to work,” he thought, “because this thing does work, but how much can we do with it?”
Rezai and his colleagues wanted to graduate to more complex and finer movements. Burkhart proposed attempting actions he wished he could perform in daily life, such as using a credit card. “I can be fairly independent when I go out to stores, but I can't hold a credit card well enough to swipe it through a reader,” Burkhart says. “Then we started working on all types of objects. Can we pick up a telephone and hold it to my ear? Can we pick up a spoon and scoop something out of a bowl and bring it to my mouth?” The researchers began working with Burkhart on a diverse range of grips and movements. The difference between a basic action such as flexing a finger and a complex task such as picking up a spoon and putting it down 10 inches away became quickly evident. “What would happen is, I would pick the spoon up, and then as soon as I would start to move my arm, it would drop out,” Burkhart says. “Being able to sustain that grip through the motion was really challenging.”
Click or tap to enlarge
Enabling these higher-order movements required a unique kind of teamwork between human and machine. Like two people from different cultures, Burkhart's motor cortex and the decoder were learning to communicate through a process of trial and error he described as a “back-and-forth game of me learning the system and the system learning me.” By early 2016 he was able to get his hand to execute acts of dexterity that months earlier had seemed out of reach: swiping a credit card, stirring a drink with a straw and even playing the video game Guitar Hero.
The Long Road Ahead
The publication of the Nature paper made headlines around the world. Weeks later Rezai and his colleagues got more good news. The U.S. Food and Drug Administration had given them permission to keep the implant in Burkhart's brain for another year. The researchers had estimated the chip would stop working properly after a year or so–the progressive scarring of brain tissue around the site was expected to gradually degrade the signal. “But it has been 700 days plus, and the signals can still be made sense of–that's unbelievable,” Rezai said at the beginning of the study's third year, although he added that the signals had grown weaker.
Nobody was more pleased about the extension than Burkhart. “I am not ready to be done with the project,” he said last October. He had driven to Ohio State's medical center–as he does two to three times a week–in a van modified for his needs. In addition to a wheelchair ramp, the vehicle has special levers that allow him to control the gas pedal, the brake and the steering wheel with his right arm, which he can still maneuver using his shoulder. (Burkhart retains the ability to use about two thirds of his shoulder muscles but has almost no functional control of either arm from the elbow down.) “I really enjoy seeing how much I can do,” Burkhart says of the sessions. He found it rewarding that “now we can do seven different movements.”
Still, the approach taken by Rezai's team has limitations. For one, Pittsburgh's Schwartz notes, the different muscles on Burkhart's arm are not controlled individually by his thoughts; rather the system selects a muscle-activation sequence from a menu of sequences, like “picking a particular tape for a player piano.” In work done at Schwartz's own lab, he says, even though subjects move a robotic arm rather than their own, “our control is elaborate enough that our subjects [can separately] operate the arm, wrist and fingers.” Burkhart's muscle control is limited, too, because the stimulating electrodes sit on the outside of his skin. In patients with a greater degree of paralysis, such external stimulation would be unlikely to work.
A group led by Robert Kirsch of Case Western Reserve University and Brown's Donoghue recently succeeded in inserting fine electrodes into the completely paralyzed arm of a patient with a brain implant, who was then able to make crude arm and hand movements.
Looking ahead, Rezai's group hopes to both expand the range of Burkhart's movements and make them more sophisticated. “When you're doing things in your day-to-day life, the strength is really important,” says Battelle computer scientist David Friedenberg. He gave the example of picking up a paper cup. “If you grip it too hard and it's empty, you're going to crush the cup. Then once you fill it up, that light grip isn't even going to lift the cup off the table,” he says. “As you're drinking it, you're constantly adjusting how much strength you're using.”
The researchers are also working to improve the decoder so it can correctly identify signals associated with a specific action without extensive training. The stimulation technology is improving as well—in recent months engineers have equipped the sleeve with sensors to keep track of changes in the position of the electrodes on Burkhart's arm as he moves.
This summer, unless the FDA grants yet another extension, Burkhart will have the implant removed. But efforts to make the technology usable will continue. Rezai envisions an implant that relays signals wirelessly and a decoder that can run on a smartphone communicating with the stimulating sleeve. In fact, researchers at Brown have already developed a wireless implant. The ideal system would not only transmit Burkhart's thoughts to the decoder but also relay back tactile feedback—as a group at Pittsburgh demonstrated in 2016, when the researchers partially restored a paralyzed man's sense of touch by electrically stimulating his brain.
“We are far from where we need to be,” Rezai says. “Ian needs to be able to take it home so that when he wakes up in the morning he can just wear a sleeve, like he wears a shirt, and he can pick up a croissant and a cup of coffee, go outside to the backyard and hang out.