Machine learning continues to be an exciting and emerging technology and this week we are finding that a machine learning algorithm could detect signs of anxiety and depression simply from the speech patterns of young children. This, of course, could potentially provide a fast and easy way to diagnose a variety of conditions that are typically hard to catch at first glance and, as such, are often overlooked in young people.
Data suggests that one in five children currently suffer from anxiety and depression. These are collectively known, medically, as “internalizing disorders.” However, since children under the age of eight are not typically reliable in articulating their emotional pangs, adults need new ways to infer their mental state in order to recognize potential mental health problems.
Lead study author Ellen McGinnis explains, “We need quick, objective tests to catch kids when they are suffering.”
Indeed, long waiting lists for appointments with psychologists, issues with insurance, and failure to recognize early symptoms all contribute to children missing out on important interventions.
The University of Vermont Medical Center Vermont Center for Children, Youth, and Families, goes on to say, “The majority of kids under eight are undiagnosed.”
Of course, early diagnosis is crucial because children tend to respond well to treatment while their brains are still in development. On the other hand, if left untreated, these children could be more vulnerable to substance abuse risk and even suicide, later in life.
The study involves the [adapted] use of a mood induction task called the Trier-Social Stress Test. This intends to cause stress and anxiety among a group of 71 children, between the ages of 3 and 8. These children were asked to improvise a 3-minute story and then told that would be judged based on how interesting their performance would be. After 90 seconds, and then with 30 seconds left, a buzzer would indicate how much time was left.
“The task is designed,” McGinnis explains, “To be stressful, and to put them in the mindset that someone was judging them.”
From here, machine learning used an algorithm to analyze statistical features in the audio recordings. Sure enough, they identified vocalizations that help distinguish internalizing disorders among children with an 80 percent accuracy.
The study has been published in the Journal of Biomedical and Health Informatics.