Case Study

Improving care for people with autism

The Challenge

Having a child or teenager with severe autism who behaves aggressively leads some families to avoid social settings, limiting their child’s ability to explore the world and diminishing quality of life for everyone involved. Northeastern Associate Professor Matthew Goodwin is seeking to develop a machine learning algorithm that, integrated with a sensing device, can predict when young people with autism are about to have an emotional outburst. By giving parents and other caregivers a few minutes’ advance warning, the device could enable them to take steps to calm the child or teen and ensure everyone’s safety.

Our research shows how different disciplines, use-inspired research, technology, academia, and industry all come together. The Roux Institute will facilitate these kinds of collaborations.

Matthew Goodwin

Associate Professor, Health Sciences and Computer Sciences

The Partnership

To measure and track autism inpatients’ aggressive behavior and associated physiological indicators, Goodwin turned to a colleague at Spring Harbor Hospital in southern Maine: Matthew Siegel, a child psychiatrist and associate professor of psychiatry and pediatrics of Tufts Medical School, who directs the hospital’s inpatient treatment center for children with autism. Siegel co-conceived their project and oversees data collection. Project funders include the Department of Defense, the Simons Foundation Autism Research Initiative, and the Nancy Lurie Marks Family Foundation.

Goodwin, Siegel, and their colleague Carla Mazefsky at the University of Pittsburgh launched a study in which inpatients with autism ages 7 to 18 wear a wristband with biosensors that monitor peripheral physiological signals—heart rate, sweat production, skin surface temperature, physical activity—and track their changes before, during, and after aggressive outbursts. At Northeastern, a team of experts in machine learning, data science, and health informatics are creating algorithms to automatically process this data. As of April, their algorithm predicted outbursts three minutes in advance with greater than 95 percent accuracy.

The Goal

The researchers say a company could commercialize a wearable biosensor that streams data to a caregiver’s phone and sends push alerts signaling an imminent outburst. But first, clinical trials will be needed to test the technology’s feasibility and efficacy in the hands of families, caregivers, and other users.

If you can dream it, we can help you build it.

To solve a challenge—or turn an idea into reality—contact us. Our expertise is broad, ranging from experiential AI, nanoscience, and coastal sustainability to supply chain management. We can help you develop new technologies and even launch new ventures.