New Research Applies Computing to Detecting Autism

• By Dian Schaffhauser
• 09/01/10

A team of researchers in eight universities has just won a five-year, $10 million grant from the National Science Foundation to create new computing techniques for measuring and analyzing the behavior of children. The goal is to create new ways to identify those at risk for autism and other developmental delays.

The team, led by the Georgia Institute of Technology, also includes scientists from Boston University, Carnegie Mellon, Emory University, Massachusetts Institute of Technology, the University of Illinois at Urbana-Champaign, the University of Pittsburgh, and the University of Southern California, as well as collaborators from several autism research centers.

The research will tap computational behavioral science, a new discipline that uses the fields of computer science and psychology to develop innovative ways to study human behavior.

Autism affects one of every 110 kids in the United States, and treatment at an early age can improve the long-term outlook for their treatment. That drives the need for a screening process that will work at an early age.

"Direct observation of a child by highly trained specialists is an important step in assessing risk for developmental disorders, but such an approach cannot be easily scaled to the large number of individuals needing evaluation and treatment," said James Rehg, lead investigator and a professor in Georgia Tech's School of Interactive Computing.

Researchers will design vision, speech, and wearable sensor technologies to analyze child behavior. Data will be collected during interactions between caregivers and children, children in a daycare environment, and clinicians interacting with children during therapy sessions.

Scientists at Emory and Georgia Tech are developing a screening process called Rapid-ABC, which uses cameras and microphones to provide an inexpensive way to measure eye gaze and facial and body expressions, along with speech and non-speech utterances. On-body sensors could measure physiological variables such as heart rate and skin conductivity, which contain clues about levels of internal stress and arousal that are linked to behavior.

Researchers from Carnegie Mellon's School of Computer Science will focus on development of the use of computer vision to automate analysis of videos for early signs of autistic behavior, such as the rocking, clapping, and other repetitive behaviors common in kids with autism.

Other researchers will focus on coming up with new methods for making sense of the data being generated by these tools.

"We hope that by incorporating this screening protocol into well-child doctor visits for children less than two years old, we can reduce the average age of autism diagnosis, which is currently about four years old," said Georgia Tech School of Interactive Computing senior research scientist Rosa Arriaga.

Rehg added that in the future the researchers expect to expand their work beyond autism to other developmental disorders and the general study of child behavior. "While autism is our focus right now, this project addresses general social, communicative and repetitive behaviors, so the technologies we develop will have applicability to other childhood disorders, such as Down syndrome or attention deficit hyperactivity disorder."