Sensor Found in "Virtually Every Wearable" Can Diagnose Childhood Anxiety
"They can't yet reliably report if or how they might be suffering."
Communicating what it feels like to suffer from anxiety is a trying task, even for the most articulate teens or adults who live with anxiety disorders. But for young children who might not have the language skills to put words to their emotions, describing that crushing feeling of panic is an impossible task. A paper published Wednesday in PLOS One proposes a solution: a wearable sensor and machine learning algorithm that can diagnose anxiety without hearing a single word. And the best part is that all the necessary parts of this technology already exist.
Mental health and anxiety disorders are already notoriously hard to pin down in teens and adults. These challenges are even greater when it comes to diagnosing children, adds Ellen McGinnis, Ph.D., a postdoctoral researcher at the University of Vermont’s psychiatry department.
“Young children are grappling with understanding their own emotions and expressive language, so they can’t yet reliably report if or how they might be suffering,” she tells Inverse.
“For instance, I tried to administer a self-report anxiety questionnaire made for children seven and up to this research sample. One of the items asked something like ‘Are you jumpy?’ and 90% of the children started jumping up and down, smiling.”
To get around this obstacle, she and study-coauthor Ryan McGinnis, Ph.D., a biomedical engineer, also at the University of Vermont (and Ellen McGinnis’s husband), reimagined a typical movement sensor found in nearly all smartphones, called a micro-electro-mechanical system — or a MEMS device. These are the nano-scaled devices, which measure acceleration and angular velocity, make up the accelerometers in “virtually every wearable and smartphone on the market,” Ryan McGinnis adds. When he strapped the MEMS device around the waists of 63 children, some of whom had clinically diagnosed anxiety disorders, he found that these children actually tended to move differently than healthy controls when they were put in stressful situations.
## The ‘Snake Task’
Unfortunately, the only way to design and test an anxiety sensor for children is to induce anxiety. Suffice to say that the snake task succeeds on this front.
A researcher leads the children into a dimly lit room, and says, “I have something to show you,” or “Let’s be quiet so it doesn’t wake up,” before pulling back a sheet to reveal a fake snake, just inches from their face. Then, the researchers allow the children to play with the snake, all the while assuring them that everything will be okay.
Children with anxiety disorders moved most differently during the first phase of the task, when researchers built suspense about what creature was dwelling behind the sheet. According to the MEMS sensor data, children with an anxiety diagnosis tended to turn away from the mysterious sheet more quickly and more dramatically than healthy controls — often completely turning their backs on it — 180 degree. Children with no anxiety diagnosis typically turned less than 60 degrees, keeping the sheet within sight.
“Many anxiety disorders are characterized by worrying about uncertainty and behaviorally avoiding uncertain situations,” explains Ellen McGinnis. “Finding that children with disorders were physically turning away fit well with psychological theory and behavioral reports of individuals with anxiety and depression avoiding potential threats.”
Screening For Anxiety
Ryan and Ellen McGinnis used this preliminary data to construct a machine learning algorithm that uses this rotational motion and speed from the REMS sensor to diagnose kids with potential anxiety disorders. So far, the algorithm can distinguish between healthy controls and kids with a diagnosis with 81 percent success. As the algorithm learns from more cases, the researchers hope that statistic will improve.
Ellen McGinnis calls this movement data an “objective measure of child anxiety” that could be used during early-life pediatrician appointments. Yet they’re not so quick to say that it could replace “gold-standard psychological interviews.” Instead, it’s intended as a supplement that might help identify children who would benefit from follow-ups with psychiatrists.
"“Young children are grappling with understanding their own emotions and expressive language, so they can’t yet reliably report if or how they might be suffering."
In that sense, this anxiety sensor and algorithm are part of a diagnostic trend. There’s evidence that algorithms are useful in at least helping flag conditions while there’s still time to intervene. The Apple Watch has already successfully done this for a heart condition, and some A.I. programs show promise for diagnosing sepsis.
Still, there are some concerns about how to classify movement data especially when it’s used in a diagnostic framework. This movement data could amount to a medical record, and Ryan McGinnis adds that it’s crucial to build privacy features “from the ground up” into the data collection process — especially given the delicate nature of an anxiety diagnosis.
“We don’t have good answers to this at this time, but our goals are to make sure all children are connected to the emotional and behavioral care they need as early as possible,” adds Ellen McGinnis. “For now, keeping this information protected within health systems, like any other doctors note, seems like a good place to start.”