McKelvey School of Engineering School of Medicine

Data from wearables could be a boon to mental health diagnosis

A team of Washington University in St. Louis researchers developed a deep-learning model called WearNet using 10 variables collected by the Fitbit activity tracker to detect depression and anxiety. (Photo: iStock)

Depression and anxiety are among the most common mental health disorders in the United States, but more than half of people struggling with the conditions are not diagnosed and treated. Hoping to find simple ways to detect such disorders, mental health professionals are considering the role of popular wearable fitness monitors in providing data that could alert wearers to potential health risks.

While the long-term feasibility of detecting such disorders with wearable technology is an open question in a large and diverse population, a team of researchers at Washington University in St. Louis showed that there is reason for optimism. They developed a deep-learning model called WearNet, in which they studied 10 variables collected by the Fitbit activity tracker. Variables included everything from total daily steps and calorie burn rates, to average heart rate and sedentary minutes. The researchers compiled Fitbit data for individuals for more than 60 days.

When considering depression and anxiety risk factors, WearNet did a better job at detecting depression and anxiety than state-of-the-art machine learning models. Further, it produced individual-level predictions of mental health outcomes, while other statistical analyses of wearable users assess correlations and risks at the group level.

“Deep learning discovers the complex associations of these variable with mental disorders,” said researcher Chenyang Lu, PhD, the Fullgraf Professor at the McKelvey School of Engineering and a professor of medicine at the School of Medicine. “Machine learning is our most powerful tool to extract these underlying relationships. Our work provided evidence, based on a large and diverse cohort, that it is possible to detect mental disorders with wearables. The next step is to convince a hospital system or some company to implement it.”

Researchers included Ruixuan Dai, who worked in Lu’s lab as a doctoral student and is now a software engineer at Google; Thomas Kannampallil, PhD, an associate professor of anesthesiology and associate chief research information officer at the School of Medicine and an associate professor of computer science and engineering at McKelvey Engineering; Seunghwan Kim, a doctoral candidate at the School of Medicine; Vera Thornton, an MD/PhD candidate at the School of Medicine; and Laura Bierut, MD, the Alumni Endowed Professor of Psychiatry at the School of Medicine.

The team presented its findings May 10 at the ACM/IEEE Conference on Internet of Things Design and Implementation. The paper was awarded the Best Paper Award for IoT Data Analytics at the conference.

Wearable data could be a boon to mental health diagnosis and treatment, according to Lu.

Read more.