Designing New Tools to Detect Early Signs of Autism in Babies

In the United States, the average age of autism diagnosis is four to five years of age, and many children are diagnosed even later. This means that many young children miss out on the opportunity for early behavioral interventions, which have been shown to improve children’s language, cognitive, and social abilities. New research finds that early signs of autism are present much earlier than when a diagnosis is typically made. In fact, for many autistic children, early signs of autism — paying less attention to people, smiling less, failing to turn when their name is called, and using fewer vocalizations — are apparent between six and 12 months of age. The first year of life is a time of rapid brain development when infants are learning about the social world and how to communicate with others. If intervention could begin as soon as early signs appear, perhaps children would have even better outcomes.

Currently, there are no valid screening tools for detecting autism in infants. A Duke Center for Autism and Brain Development team of investigators led by Geraldine Dawson, PhD, and Guillermo Sapiro, PhD, received funding from the National Institute of Mental Health to create and validate an autism screening tool that would be appropriate for infants as young as 6 months of age. They set out to develop a tool that was easy to use by both parents and pediatricians and could accurately and reliably detect early behavioral signs of autism that are seen in infants. To achieve their goal, they designed a digital application (app) that parents can download on their smart phone or tablet at home.

The app, which is available in English and Spanish, consists of brief, developmentally appropriate, specially designed movies that are shown to the infant on the parent’s device while sitting on their parent’s lap. The infant’s responses tothe movies — including their eye gaze, facial expressions, body movements, and vocalizations — are recorded by the camera in the device and then automatically coded using artificial intelligence and computer vision. Using these methods, the data collected shows whether the child is more interested in the social or non-social elements shown in the movies, and tracks their emotional responses and other behavioral signs of autism. Once the infants are 18–24 months of age, psychologists on the team offer a free diagnostic evaluation to any toddler for whom there is a developmental concern. The research project, called the Sense to Know (S2K) Study, is being conducted in collaboration with Duke Primary Care providers.

“We have enrolled more than 1,000 infants already, and early results are promising that this novel and scalable approach will offer accurate, reliable screening for autism in infants,” said Brian Eichner, MD, a Duke Primary Care pediatrician and assistant professor of pediatrics. “Our work could lead to an easy-to-use, objective diagnostic tool for doctors and parents, and could mean that kids get access to provided therapies earlier than most do now.”