Tool Diagnoses Autism Through Children's Gaze

Clinical studies show that measuring children's looking behaviour predicts a diagnosis of autism in children aged 16 to 30 months with high accuracy.

According to researchers from Marcus Autism Centre, a division of Children's Healthcare of Atlanta, this new tool can assist clinicians in diagnosing autism earlier while also providing objective measurements of each child's strengths and vulnerabilities, allowing them to jumpstart effective support for children and families.

The findings of the study were published in the Journal of the American Medical Association (JAMA) and JAMA Network Open.

 “The results show that the way in which young children look at social information can serve as an effective and objective biomarker for early signs of autism,” says Warren Jones, PhD, lead author, Director of Research at Marcus Autism Center at Children’s Healthcare of Atlanta and Nien Distinguished Chair in Autism at Emory University School of Medicine.

In early trials of the device, researchers frequently referred to the biomarker technology as “the Marcus Test” for autism, acknowledging the leadership of philanthropist Bernie Marcus in giving autism research and treatment over the past 35 years, including the founding of the Marcus Autism Center.

One of the crowning achievements of his life’s work in autism has been the development of the tool to diagnose children earlier and get them into treatment sooner.

Autism affects 1 in 36 children, meaning that each year, in the US alone, more than 90,000 children with autism are born. Early identification and early intervention are important for supporting the health, learning, and long-term well-being of all children with autism.

For more than two decades, Dr Jones and co-author Ami Klin, PhD, Director of the Marcus Autism Center at Children’s Healthcare of Atlanta, have studied "social visual engagement”—how children look at and learn from their surrounding social environment—and how this differs in children with autism.

Past research showed that these differences emerge early in infancy and are directly related to individual genetic differences. In their current work, Drs. Jones and Klin developed technology to reliably measure these differences as a biomarker for clinicians to use.

“The far-reaching implications of these results may mean that children who currently have limited access to expert care, and face two or more years of waiting and referrals before finally being diagnosed at age four or five, may now be eligible for diagnosis between the ages of 16 and 30 months,” said Dr Klin, who is also Division Chief of Autism and Developmental Disabilities at Emory University School of Medicine.

“In addition, this technology measures each child’s individual levels of social disability, verbal ability and non-verbal learning ability, which is critical information for clinicians when developing personalized treatment plans to help each child make the greatest gains.”

Research published in JAMA Network Open describes the initial development of the technology, with results from more than 1,000 two-year-old children in discovery and replication studies.

Data published in the Journal of the American Medical Association (JAMA) describe results from a multi-site trial at six of the nation’s leading centres for autism diagnosis and treatment, using automated devices to test approximately 500 children between the ages of 16 to 30 months.

In each of these studies, the researchers tested whether automated measures of children’s looking behaviour could effectively predict the results of diagnostic assessments conducted by highly trained, expert clinicians.

During testing, children watched video scenes of social interaction. As they watched, their eye movements were monitored at a rate of 120 times per second to determine, moment-by-moment, what social information the children looked at and what they did not.

After collection, tens of thousands of these measurements were compared to data from age-matched peers, using algorithms to quantify similarities and differences at each moment in time.

These measurements were summarized to provide an overall diagnostic indication as well as individual measurements of each child’s levels of social disability, verbal ability, and non-verbal learning skills.

When compared with an expert clinical diagnosis of autism at six of the nation’s leading centres, automated measurements of children’s looking behaviour accurately matched the current gold standard, which could help pave the way for earlier, objective diagnosis in many children.

“Objective, performance-based biomarkers can help clinicians diagnose and support more children and families, with the same level of clinical confidence. We hope this can alleviate current burdens on the healthcare system and reduce lengthy waitlists for assessment,” said Dr Jones. “When we can shorten the time from parents’ first concern to diagnosis and the start of beneficial support, we can help positively impact the lives of many children and families.”