Support Provided ByLearn More
Body + BrainBody & Brain

Artificial Intelligence Detects Early Signs of Autism in Infants

ByAnnette ChoiNOVA NextNOVA Next

Receive emails about upcoming NOVA programs and related content, as well as featured reporting about current events through a science lens.

By using artificial intelligence to decipher brain scans, researchers were able to predict with startling accuracy which high-risk infants ended up developing autism by age two.

Detecting autism before age two has been notoriously difficult, in part because the disease is diagnosed based on behavioral traits, including a difficulty to make eye contact or respond to their name. Many of the telltale signs aren’t present in children that young, especially infants younger than one year old.

Support Provided ByLearn More

But the study, reported in the journal Nature , suggests that identifying autism spectrum disorder (ASD) before symptoms arise may not be as far-fetched as scientists once thought.

screen-shot-2017-02-23-at-2-02-55-pm
The colored cortical areas represent regions with significant expansion of surface area from six to 12 months.

Though doctors who monitor high-risk children with ASD currently do not intervene until clinical signs appear, researchers believe the brain changes that come with ASD develop much earlier, possibly as early as in the womb.

According to the Center for Disease Control, approximately one in 68 children have ASD. However, infants with a diagnosed older sibling have a one in five chance of developing ASD. The new study was funded by the U.S. National Institutes of Health and scanned 106 high-risk infants at six, 12, and 24 months.

At 24 months, 15 of the 106 babies were diagnosed with ASD. In the MRI results, researchers noticed that their brain volume grew more quickly in the 12 to 24 month period compared with those without ASD. During this time, behavioral symptoms of autism began to surface as well.

But detectable changes started happening even earlier. In those same infants, between the six and 12 months, researchers also found rapid growth of the surface area of the cortex, which are the outside folds of the brain. While researchers have known since the 1990s that children with autism can have larger brains, suggesting it may be a telltale sign of the disorder, the timing and direct correlation of the enlargement to ASD behavioral symptoms have not been confirmed.

The AI analysis of the MRI scans at six and 12 months allowed the researchers to forecast, with 81% accuracy, which high-risk infants would develop autism by age two.

Here’s Delthia Ricks, reporting for Newsday:

Team members in the study, led by medical investigators at the University of North Carolina at Chapel Hill and including 10 other institutions, credited their predictions’ accuracy to a customized algorithm that was applied to the data in classifying children most likely to meet criteria for autism by age 2.

Dr. Joseph Piven, the study’s senior author, said the findings open a new window into the diagnosis of autism.

“Typically, the earliest an autism diagnosis can be made is between ages 2 and 3,” he said. “But for babies with older autistic siblings, our imaging approach may help predict during the first year of life which babies are most likely to receive an autism diagnosis at 24 months.”

The ability to diagnose autism through brain scans could potentially allow doctors to identify which patients would benefit from early intervention. Today, most efforts are directed at younger siblings of children diagnosed with autism, but not all of them go on to develop the disorder. Being able to more accurately target infants for early intervention would direct resources to where they’re most needed.

Though the current study focused on siblings, researchers agree that a larger follow-up study of high-risk infants should be conducted to confirm the results and test whether ASD can be predicted in the general population as well.

While there is no evidence that the risk of developing autism can be reduced in infants, an early diagnosis would be informative to not only test interventions, but also develop a better understanding of effective treatments.

Photo credit: Piven et al. 2017