UVM researchers study signs of childhood depression in speech

BURLINGTON, Vt. (WCAX) About one in every five children suffer from anxiety or depression. Many go undiagnosed because they are too young to express themselves and adults miss the symptoms. New research from the University of Vermont found machine learning algorithms could detect signs of those disorders in young children just by looking at their speech.

"Okay Spencer, we're going to do an activity now," said Ellen McGinnis, who along with her husband Ryan, are researchers in UVM's biomedical engineering program.

Spencer Bryant's activity is to tell a story, any story, for three minutes. It's a tough task for an energetic three-year-old.

"People are going to watch you and judge you based on how interesting your story is, okay?" McGinnis tells Spencer.

The exercise puts him under a bit of stress to come up with something to talk about.

"Once upon a time there was a lion and a tiger and they were biting each other," said Spencer, beginning his story.

Adding to the stress is a buzzer telling Spencer his time is running out.

Reporter Cat Viglienzoni: Why do the buzzers?
Ryan McGinnis: Yeah, so we do the buzzers to do sort of a surprising stimulus for the child.

That's because the researchers aren't actually judging his story, at least not the content. They're looking at his voice patterns instead, seeing how he responds to pressure. Ryan showed us Spencer's first test. "You can see he reacts, like 'What is that,' and actually walks to see what happened," Ryan explained.

After the test is done, the McGinnises will take the audio and look for certain indicators in Spencer's voice that may point to anxiety or depression. Their work found that kids with those diagnoses spoke with more monotone voices and spoke in a higher pitch after the buzzer startled them. "For this study in particular, we just focused on voice, and in a way, that almost makes it easier to deploy because it's really easy to record voice," Ryan said.

Ellen McGinnis, a clinical psychologist, says the goal is to create an objective, short test to assess the mental health of kids under the age of eight. "Kids that young can't express their own emotions, so they can't reliably tell us if they're suffering or how they're feeling," she said.

She says about 20 percent of kids have an anxiety disorder and about two percent of kids have a depression disorder in preschool. "So, if left unchecked, these are going to continue," Ellen said.
"Being able to catch them and address them really early with family therapy could be extremely helpful."

They're still several years away from being able to put out a mobile app that would be used by pediatricians or schools to screen children. Their next step is to replicate their findings in another, larger group like Spencer's. Though for the moment, his work is done.

The McGinnises' says their method is not a diagnostic test, but rather a screening tool. For a child to actually be diagnosed with anxiety or depression they would need to go through a more thorough assessment with trained clinicians. But the goal is to better identify which children might need to go through that more detailed process.