Algorithm predicts autism diagnosis in young children with 81 percent accuracy

Why it matters to you

An algorithm that’s able to accurately predict autism diagnoses in young kids could enable potental interventions to be made earlier.

A team of researchers at the University of North Carolina at Chapel Hill have developed a deep learning algorithm that can accurately predict whether a child at high risk of autism is likely to be diagnosed with the disorder in early childhood.

The algorithm was able to predict with 81 percent accuracy whether a diagnosis of autism would be made for a child with an autistic sibling,. That’s considerably higher than the 50 percent accuracy of current behavioral questionnaires.

More: Friendly educational robot designed to help kids with autism

The deep learning tool was developed in conjunction with computer scientists from the College of Charleston as part of the Infant Brain Imaging Study, which focuses on early brain development among children with autism. By scanning their brains at 6 months old,  a year old, and 2 years old, they were able to make some interesting discoveries.

“In previous literature, we’ve found brain volume enlargement in autism, meaning that people with autism have bigger brains than average,” senior author Dr. Heather Hazlett, a psychologist and brain development researcher, told Digital Trends. “In this study, we add to that by pinpointing that it’s really during the first two years of life that we see this change happening. What we found is that it occurs between 12 and 24 months. It’s not present at 12 months, but emerges rapidly after that, during the second year of life. However, what we see during the first year is surface area enlargement, referring to the folding outer contour of the brain. In children who have autism at age 2, there’s a hyperexpansion or rapid growth of the surface area at 6 to 12 months. This precedes the brain volume enlargement, and is really a story that hasn’t been told before.”

Using information concerning brain surface area, brain volume, and the insight that boys are more likely than girls to develop autism, the algorithm was able to identify eight out of 10 kids with autism.

So how could this research improve the lives of those who are diagnosed?

“I think this has potential in the sense that you could target a time period early in development which might be presymptomatic,” Hazlett said. “What we believe is that the earlier you can intervene, before the behaviors and brain differences have consolidated, you may have the greatest chance to make a change in that trajectory. Intervention that could be made prior to the onset of autism, aged 2, may lead to a greater effect, since the brain is very malleable during that period.

Related posts

Leave a Comment