Mega Doctor News
By Kristen Munson
Newswise — Right now, your smartphone—part texting device, part camera, mostly digital oracle—is collecting data. Where you go. The number of steps it takes to get there. Elevation climbed. Your phone listens for you to speak to Siri, the angel of search. Data is gathered as we traipse around the Internet, browsing and clicking and googling, inadvertently dropping cookie crumbs behind. Big data adds up. But how can it all become individually useful?
For Bryn Loftness, a doctoral student in the University of Vermont’s Complex Systems and Data Science program, the answer involves getting personal with it. Loftness studies wearable technologies and develops algorithms that can improve human health. In late March, the National Science Foundation (NSF) awarded her a $147,000 Graduate Research Fellowship, distributed over the next three years, for her work uncovering the digital phenotypes (think digital fingerprints) for childhood internalizing disorders like anxiety and depression.
At 24, Loftness is a digital native. Before she could reach the keyboard on tiptoe, she recalls “dinking around” on the family computers trying to unlock their special powers. “I remember how fun it was to experiment with them,” she explains in the break room of UVM’s Mass Mutual Complex Systems Center in Innovation Hall.
Her virtual sandbox was Microsoft Word and Publisher where she once listed and alphabetically sorted all the movies in her family’s collection. As Loftness grew, she wanted computers to do more. Today, she wears that desire in the form of a smartwatch strapped to her wrist, which tracks her sleep and, until February, her steps.
“I was on a 778-day, 10,000 step streak up until one day when I was working too late and I was accidentally 500 steps too short,” she confesses. “It did feel like the end of an era.”
But she realized the metric itself didn’t matter. Her goal ultimately was to lower her resting heart rate and stay active.
“It’s great to have data,” Loftness says. “But until you do something with it, it doesn’t really matter.”
Like track the spread of a novel respiratory virus through human waste products.
Loftness recalls the early days of the COVID-19 pandemic. It was late March of her senior year in college—her birthday week. People were suddenly locked down. Classes nationwide were flipping to online formats. Loftness, then an undergraduate at Colorado Mesa University, was majoring in computer science and developing stress detector biosensors. Her focus quickly shifted to studying poop.
“I am not one to just sit idly by,” she says.
Loftness knew data could help chart a course through the pandemic. One strategy her university used involved measuring coronavirus levels present in campus wastewater to determine the location of COVID outbreaks. Loftness helped the school’s wastewater team build the instrument.
Now, in her second year at UVM, Loftness is refining another engineered health tool to identify other adverse conditions that often go undetected: anxiety and depression in young children. While nearly 20 percent of children are affected by internalizing disorders, diagnosing them is complicated. Young children may not have the vocabulary or ability to communicate problems to caregivers who might be able to help. But what if a simple device could potentially flag cases early on?
That’s the idea behind Loftness’s project, “Discovering Digital Phenotypes of Childhood Internalizing Disorders for Point-of-Care Diagnostics.” Her work builds on that of her faculty mentors at UVM, Ryan McGinnis, Karl and Mary Fessenden Professor of Biomedical Engineering, and Ellen McGinnis, assistant professor of psychiatry, to identify the biomarkers associated with internalizing orders in young children. Biomarkers are biological signatures of a disease such as an enzyme linked with a type of cancer.
Determining biomarkers for mental health issues in kids is a problem the McGinnis’ have tackled since they were doctoral students at the University of Michigan. At the time, Ellen was studying the intergenerational transmission of risk—the idea that a mother who experienced trauma in childhood may encode a stress response in her own children. During a therapy session with a five-year-old McGinnis suspected was depressed, she learned there were no existing tools to identify anxiety disorders in young kids.
The field of psychology has generally not examined the mental health of young children. It hasn’t really considered anxiety and depression something that young children experience, McGinnis explains. “They’ve been largely ignored until about age eight.”
And the symptoms of anxiety and depression in children can be very different from those in adults, often making them harder to pinpoint. Diagnosing children often involves relying on reports from the adults who see them most and in different contexts—their caregivers and teachers—McGinnis says. “Even parents in the same household see really different things. … Maybe there is something in addition that we can get from their physiology and their biology that tells us ‘wow, they are reacting to that stimulus, that stressor, in an exaggerated way.’”
Because right now, young kids struggling with anxiety and depression all too easily fly under the radar.
“We need tools to get at what’s going on with these kids,” McGinnis says. “And in the research world, that is putting them through some staged situations and seeing how their bodies react.”
McGinnis tested 75 kids in experiments designed to trigger an emotional response like approaching a stuffed snake in a dark room, or giving an impromptu speech. She spent three years tediously coding videos of the subjects’ movements with an “army of undergraduates.” A step back would be coded one way. A facial expression another. Learning to code for all the various responses took time and training.
“It was highly inefficient,” McGinnis explains.
One night, her husband, Ryan, who was researching wearable devices to measure athletic performance, suggested capturing these experiments in a much simpler way using sensors and machine learning—a form of artificial intelligence that can help with pattern recognition. Over Thai food, the idea of a wearable sensor-based system was born. Now, they just had to build it. And that is where Loftness comes in. She is the couple’s first joint graduate student, and her research will help move the project out of the lab and into doctors’ offices.
“It’s different from a lot of academic research because we are thinking of the how to translate the technology from the beginning,” McGinnis explains. “Bryn has been amazing and is forging that path with this tool.”
Over the years, researchers in UVM’s M-Sense Research Group, which Ryan directs, have flagged potential indicators of internalizing disorders using data collected from multiple sensors that track subjects vocal patterns and physical movements as they responded to stimuli, both positive (playing with bubbles!) and negative (walking towards a fake snake!). The key is finding patterns that correlate with anxiety and depressive disorders. In 2021, Ellen McGinnis was awarded an National Institute of Mental Health K23 Career Development award, and Ryan McGinnis was awarded an NSF CAREER award to support this work. Loftness has added physiological response measures like skin temperature and electrocardiography to the algorithm and aims to refine the tool to a simplified sensor that can connect wirelessly with a smart device and be deployed at scale.
“We have done multiple studies on each of these different behavioral tasks or mood induction tasks, and we are starting to see these similar results,” she explains. “And while we still need to validate findings in more diverse populations, it is starting to help our understanding of what is the digital phenotype, or digital fingerprint, of these different disorders.”
The digital system is also validated against standard diagnostic tools—the traditional method of having a psychologist interview the parent about their child’s symptoms and decide if they meet threshold for a disorder. A system equipped with wearable sensors and machine learning can analyze multiple facets of a patient response in seconds. In Vermont, the current waitlist time for getting tested by a clinician is nearly one year, Loftness explains.
“My whole goal coming into this Ph.D. was not to come out an academic professor or go into industry and, you know, ‘work for the man,’” she says. “It was to build a toolkit or create some sort of software that we could commercialize and bring to the community… that is actually what we are doing now.”
In November, she and the McGinnis’ completed an NSF regional I-Corp entrepreneurship training program and in the spring, she and Ellen McGinnis was awarded $50,000 by the NSF to develop a business plan and evaluate product-market fit. Loftness spent the spring interviewing pediatricians, hospital and clinic administrators, and parents of preschool aged children to make sure the device is not just accurate, but easy to use. Ideally, the assessment will be performed in five minutes and begun as soon as a patient is checked into the doctor’s office. Physicians often have just minutes to comb through a year’s worth of developmental data, she explains.
But where does Loftness see this work going?
“Just like you have height charts and weight charts of a child’s trajectory of growth over time, I’d like to see behavioral health over time,” she says.
That information would alert parents and health care providers that a child may be at risk of developing an anxiety disorder, and connect them to services early on.
“We are not trying to medicate kids here,” Loftness says. “It’s just to say, ‘here is where you are at and here is what we can do about it.’”
Another benefit of this type of screening device is the way it links mental health to physical movements. For McGinnis, this may help de-stigmatize seeking help for these conditions. Caregivers will be able to see that a child is affected by something emotionally and it’s showing up physically, she says.
It may also help counter the idea that ‘kids are resilient.’
“Kids are physically resilient,” McGinnis says. “They break bones, they get better. We actually know from a lot of research that they are more vulnerable than adults when it comes to mental health because they haven’t had the practice of building up those [coping] tools. They get PTSD at higher rates than adults after a trauma.
“There is so much resiliency in children that we should give them credit for,” she says. “But mental health is kind of a different game that you need to learn and practice coping strategies.
Rates of adolescent anxiety and depression have steadily risen in the United States since even before the pandemic. Perhaps flagging children earlier will help drive cases down by giving caregivers of children the information to them develop lifelong coping skills.
“At UVM, I see a lot of adolescents,” McGinnis says. “And I will ask ‘how long have these symptoms been going on for?’ And so many of them say ‘for as long as I can remember.’ … Adolescence is hard for everyone. We know the vulnerability of depression skyrockets from 13 to 15, especially for girls. Let’s give them those tools and practice before they get to that really vulnerable part.”
But first, that means collecting, sifting, and sorting more data. And that is exactly what Loftness wants to do.
“Some people have called digital mental health apps the Wild West,” she says. “But I think data itself is the Wild West in how we utilize it. Just like with the wastewater system when the pandemic hit. We can choose to be scared, or we can choose to make a system to track our poop. We can be scared of data, or we can use it to better understand how we can live better lives.”
