New Technology Uses AI and Machine Learning to Identify Pediatric Genetic Disorders


By: Chris Hayhurst

September 3, 2021

Categories: AAMI News, News Types

For children with certain rare genetic disorders, the difference between life and death too often depends on their diagnosis. If they’re lucky enough to be identified early, the interventions they have available can be more effective, and developmental support can be provided when it’s needed most. When they’re missed, on the other hand, that care may never come—or, when it does, it may be too late.

 Pic

 It’s that bleak reality that’s kept at least one researcher—Marius George Linguraru, D.Phil., M.A., M.Sc., of Children’s National Hospital in Washington, DC—focused on his work in recent years. Principal investigator at the hospital’s Sheikh Zayed Institute for Pediatric Surgical Innovation, Linguraru has, since 2012, led a multidisciplinary team of scientists attempting to develop software that would simplify screening for such diseases. Their cutting-edge research involves the use of patented software that leverages artificial intelligence and machine learning to analyze biometric data and identify facial markers that are indicative of genetic disorders. And now, Linguraru says, they’ve reached a milestone in that work: They just signed a licensing agreement with a medical device company—a startup called MGeneRx Inc.—to commercialize the technology and put it in clinicians’ hands.

“The goal is to develop it to the point where it’s like having a specialist in your pocket,” Linguraru explains. He won’t speculate how MGeneRx might package the software for clinical use, but in his team’s research they deployed the technology using a smartphone app. “You capture the data by taking a picture, and it’s processed automatically,” he says.

Earlier Detection, Better Outcomes

The hospital’s collaboration with MGeneRx was arranged by the Children’s National Office of Innovation Ventures, a division created specifically to advance the organization’s research and commercialize its inventions. The Children’s National Rare Disease Institute also played a role in the technology’s development, and the research was informed by a number of studies conducted in conjunction with the National Human Genome Research Institute at the National Institutes of Health. Their findings, according to Linguraru, indicate that their software has the potential to detect at least 128 genetic diseases with 88% accuracy. These diseases include rare but relatively well-known conditions like Down syndrome, Williams-Beuren syndrome, and Noonan and DiGeorge syndromes.

Linguraru and his counterpart at MGeneRx, acting chief executive officer Nasser Hassan, agree that the tool should prove particularly useful in regions of the world with limited access to geneticists and genetic testing.

“The social impact of this technology cannot be underestimated,” Hassan said, in a statement released when the licensing agreement was announced. That’s true in countries everywhere, he added, but it’s especially the case in low-income nations, “where there is a high prevalence of rare genetic conditions but a severe lack in the specialty care required to diagnose and treat them.”

Down syndrome provides one good example of how early intervention can make a difference. “We’ve made extraordinary progress for the care of children and adults with Down syndrome over the years, mainly because we’ve learned to identify it sooner,” Linguraru says. Back in the 1960s, the life expectancy of a newborn with Down syndrome was about 10 years, he notes. Today, more than 6,000 babies are born with the chromosomal disorder each year, and statistics show they can expect to live a functional life well into adulthood. “It’s not that we have a magic wand that makes the condition go away, but many of the associated risks that come with the condition are handled earlier and better,” Linguraru says.

“Distilling the Clinician’s Brain”

In their work at the Sheikh Zayed Institute, Linguraru and his team looked to other facial-recognition technologies for their initial inspiration, he says. “Facial analysis has been used in a variety of applications—cybersecurity, airport security; it’s becoming a lot more common all over the world. At the same time, when it comes to children with genetic conditions, we’ve always known that they’re often identified by their primary care physicians, or when they’re seen by a specialist, or sometimes by members of their families who may note there’s something different about the pattern of their child’s face.” Geneticists and morphologists have studied these facial characteristics in depth, Linguraru says, but few clinicians have the experience required to absorb and apply that knowledge themselves.

“What’s really difficult to learn is the huge variety of these patterns, because that can only come with time,” Linguraru explains. A physician who sees a large and diverse population of patients might recognize the signs of a rare genetic condition because they’ve seen them before. But anyone else—unless they’re a specialist—could easily assume the child appears the way they do because of normal genetic variation. “If you’re a physician who hasn’t been trained to identify dysmorphology, you may not notice anything is wrong,” he says.

Their technology, Linguraru explains, scans, and analyzes the biometric data points of a child’s face in real time. And because it relies on AI algorithms for the processing, the more data it’s able to capture with each successive patient, the more effective it potentially becomes. “That is one of the great advantages of artificial intelligence—the algorithms can get smarter and smarter over time. But it’s just like the clinician who needs to see diverse kids to understand where a change in facial morphology is natural and due to ethnic variability versus pathological and caused by some chromosomal condition. Access to data is key, and not just to a large amount of data, but data that is diverse within the population.”

Every step of the way in the technology’s development, he and his colleagues worked closely with clinicians at Children’s National Hospital to ensure it was designed to meet their needs, Linguraru says. “I like to think that in my line of research, we get to distill the clinician’s brain. We pick up the way they think, the way they look at their patients, and we turn that into computer code and artificial intelligence approaches that work in a similar way.” In this case, with their biometric analysis technology, he’ll call it a success if the final product is not only effective, but “portable, affordable, and reproduceable.” And critically, he says, if clinicians are to use it, it should just become another tool in the toolbox they have available for helping their patients get the care that they need.

“Technology like this is great and has amazing potential for bringing opportunities to patients who may not have equitable access to elite clinical care, but identifying those patients is just the first step,” Linguraru says. “To really have an impact on health outcomes, you have to ensure they’re taken care of within the healthcare system, and within their society.”

The acceleration and commercialization of this objective digital biometric analysis technology will not only help diagnose rare genetic disorders—it will also allow for earlier interventions that improve the quality of life for the children living with these conditions.”

Kolaleh Eskandanian, PhD, MBA, PMP, vice president and chief innovation officer, Children’s National Hospital, and managing director, Innovation Ventures