Consumer Electronics

Machine learning algorithm can detect anxiety and depression in children

10 May 2019

New algorithm can diagnose children with an internalized disorder within a few minutes. New algorithm can diagnose children with an internalized disorder within a few minutes.

Researchers from the University of Vermont (UVM) created a machine learning algorithm that can detect internalized disorders based on a child’s speech. The algorithm could provide a fast and easy way for doctors to diagnose difficult-to-detect disorders in children.

Internalized disorders are anxiety and depression. One in five children suffer from anxiety and depression, but many parents do not know their child is suffering until they are much older. Early diagnosis is important because children with these disorders respond well to treatment. If left untreated, these children have a greater risk of substance abuse and suicide later in life.

Currently, diagnosing internalized disorders is a long process. The child sits in a room with their primary caregiver and a clinician. The physician conducts a semi-structured 60- to 90-minute interview with the child and then determines the state of the child’s mental health. The UVM team wanted to create a quick and simple tool that streamlines the diagnostic process within a few minutes.

While testing the algorithm, the team used an adapted version of the Trier-Social Stress Task. The task is designed to cause feelings of stress and anxiety in the participant.

Participants in the study included 71 children aged three to eight. The children had been previously diagnosed with an internalized disorder through a clinical interview and parent questionnaire. The children were asked to improvise a three-minute story while a researcher listened and acted as a stern judge. The judge only gave neutral or negative feedback. The children were told that they would be judged by how interesting their story was. A buzzer rang after 90 seconds and then again when only 30 seconds were left. The buzzer was designed to be stressful for the child.

After gathering recordings of the stories, the team fed the audio to the new algorithm. The algorithm analyzed the statistical features of the audio and related the features to the child’s diagnosis. The algorithm diagnosed the children with 80% accuracy in only a few seconds. The team found that the middle part of the recording, between the two buzzers, was the most revealing of the child’s disorder.

To reach a diagnosis, the algorithm identifies eight features of speech, three of which are linked to internalizing disorders. These features are a low-pitched voice, a higher-pitched response to the buzzer and repeatable speech inflection and content. The study also found that children with internalizing disorders showed a larger turning-away-response from fear stimulus.

The UVM team believes that voice analysis could be revolutionary for diagnosis in a clinical setting. But the algorithm still needs further development, including turning the algorithm into a smartphone app. The researchers also hope that they can use the algorithm to create a universal screening tool for mental disorders.

A paper on this technology was published in the Journal of Biomedical and Health Informatics.



Powered by CR4, the Engineering Community

Discussion – 0 comments

By posting a comment you confirm that you have read and accept our Posting Rules and Terms of Use.
Engineering Newsletter Signup
Get the GlobalSpec
Stay up to date on:
Features the top stories, latest news, charts, insights and more on the end-to-end electronics value chain.
Advertisement
Weekly Newsletter
Get news, research, and analysis
on the Electronics industry in your
inbox every week - for FREE
Sign up for our FREE eNewsletter
Advertisement