Looking concussion in the eye
Ask three neurologists if someone has a concussion, says Rosina Samadani ’89, SM ’92, and you’ll get four different opinions. A concussion, typically caused by a blow to the head, is an injury that temporarily impairs brain function. But the damage usually doesn’t show up on brain scans—and with no clinical standard, or even an agreed-upon symptom checklist, diagnosis comes down to a judgment call based on hard-to-assess things like how nauseated or drowsy someone feels. Samadani and her sister, Uzma Samadani, an associate professor of neurosurgery at the University of Minnesota, want to change that.
About seven years ago, Uzma was trying to determine whether severely brain-injured patients were improving, so she developed an eye-tracking device to see if they could follow a moving video on a screen. The practice of assessing eye movement to detect brain injury dates back thousands of years, but it largely died out with the advent of radiology and CT scanning. “As soon as radiology came along and you had these beautiful pictures of the brain,” Uzma says, “people stopped relying so much on looking at physical eye movements.”
As she evaluated the data gathered by her device, she realized it was allowing her to detect not only the patients’ tracking ability but also restrictions in their eyes’ range of motion. It dawned on her that these restrictions indicated problems with specific nerve pathways, meaning she had found a way to measure—and possibly locate—brain injury.
Eyeball and pupil movement are controlled by three pairs of cranial nerves that emerge from the brain stem and extend forward, one nerve along each side of the brain, to connect with the eye muscles. Pressure on those nerves slows their activity. Since each cranial nerve pair governs different aspects of eye movement, Uzma reasoned, eye-tracking data could reveal which nerves are affected by pressure or damage in the brain, and even point to the site of an injury. For example, if movement is impaired in both eyes but the eyes are coordinated, that likely means the injury is in the brain stem, which would affect both sides of the nerve pair. If only one eye is affected, that suggests the injury is in a section of nerve that has already emerged from the brain stem.
Uzma planned to make the technology available free online, but she realized that few facilities would have the resources to properly set it up. So she founded Oculogica in 2013 to sell it in ready-to-use form. She got input on starting the business from Rosina, who’d studied mechanical engineering at MIT, earned a PhD in biomedical engineering from Northwestern University, and had 18 years of experience in the medical-device industry as a health-care consultant (including seven years at McKinsey) and founder of two startups. The sisters had never imagined they would collaborate professionally, but Uzma thought Rosina had the perfect skill set for Oculogica; she hired her as CEO in 2015.
The company’s tabletop eye-tracking device, a machine called EyeBox, sends a small video clockwise around the perimeter of a rectangular screen for 220 seconds. (The test videos, which include a Shakira music video, were carefully vetted to meet a list of about a dozen criteria; for example, they can’t have flashing lights.) As a patient watches the video from a stable headrest, a binocular camera separately tracks each eye and gathers about 100,000 data points at high frequency. The data feed into algorithms that calculate nearly 100 different metrics quantifying such things as speed, coordination, and range of motion. Using statistical analysis and machine learning, Uzma identified the metrics most strongly correlated with concussion in clinical studies, and she developed an algorithm based on those metrics to score the severity of a brain injury. A score of 10 or above is Oculogica’s threshold for concussion. (Uzma has seen scores as high as the low 30s; Rosina says she usually scores 1.5 to 2.5 in her normal, healthy state.) EyeBox also detects intracranial swelling, which can be caused by a concussion and other brain conditions.
While other medical eye trackers are on the market, those require a baseline and assess only attention—essentially, the patient’s willingness to follow a chosen stimulus, assuming the capacity to do so. EyeBox, which doesn’t require baseline testing, measures the function of the cranial nerves, providing what has been so elusive in brain injury medicine: a physiologic indicator of brain function. This type of objective assessment could lead to a way of classifying brain injuries and tracking recovery. The hope, Rosina says, is that it would “open the floodgates on developing appropriate therapies.”
EyeBox is in clinical trials, and Oculogica is working on getting FDA clearance for concussion diagnosis—something no other device currently has. The Department of Defense, the United States Olympic Training Center, and two high schools in Beaver Dam, Wisconsin, are among the organizations that have tested the device. Other EyeBox studies are under way at Boston Children’s Hospital, the Mayo Clinic, and Children’s Hospital of Philadelphia.
As Oculogica awaits the results of those studies, Uzma continues to identify new metrics and refine the device’s algorithms, and the company is developing a more portable version of EyeBox. A smaller version could be used in the field by the military, or possibly even on the sidelines by coaches and trainers.
Beaver Dam High School’s head football coach, Steve Kuenzi, says the only approved test now available to coaches requires kids to answer questions on a computer, both before an injury and after. “I’m not a big believer in what it tells you,” Kuenzi says, explaining that you can’t know how much effort a student puts into such a test. Studies also show that baseline tests can be unreliable: athletes can purposely do poorly to increase the odds they’ll be able to keep playing after an injury.
When Rosina tested EyeBox on about 100 of Kuenzi’s players, she approached him about retesting several who’d received high scores. First on her list was a player who had been hit hard in practice the night before but passed a traditional concussion assessment. Another was a student who had been injured in a game a few days earlier; most of the others played in “high-collision” positions.
“The first kid she mentioned to me, it kind of sold me,” Kuenzi says. The fact that a player who had just suffered a major hit had the highest score—above 10—“was pretty telling,” he says. Another student tested low but then got hit in practice that afternoon; when Rosina tested him with EyeBox again at the hospital that night, his score jumped significantly, says his mother, Kelly Braker. He was tested repeatedly over the following months, and Braker was reassured as his score gradually declined. “You’re like, okay, obviously he’s getting better,” she says.
By minimizing the guesswork involved in diagnosing concussions, Kuenzi says, that kind of tool can be a big help to coaches trying to protect players and defend a sport increasingly seen as too dangerous. “You’re talking the brain,” he says. “We’re not gonna mess around with that.”
AI is here.
Own what happens next at EmTech Digital 2019.