As an NIH-funded Autism Center of Excellence, the researchers' data and tools are open-source and will eventually be submitted to the NIH's National Database for Autism Research.
Hazlett and Piven's team then used a deep-learning neural network, a form of machine learning, to ask if MRI scans at 6 and 12 months in a larger set of high-risk infants could predict an autism diagnosis at age 2. But children in groups with higher rates of autism - like those who have affected siblings or who carry genes linked to autism - might benefit the most from it.
If the algorithm predicted that the child would be diagnosed with autism, it was correct just over 80 percent of the time.
Jed Elison, assistant professor of pediatrics at the U of M, co-authored a paper on the research published Wednesday in the scientific journal Nature, and said the results shed new light on autism. The study's lead site was based at University of North Carolina-Chapel Hill.
But the diagnostic breakthrough addresses a key problem that has confounded efforts to effectively screen for autism as quickly as possible: Babies typically don't show clear outward signs of the disorder until almost the end of their second year of life.
But if the study results are confirmed in future research, it could offer a new option for screening high-risk children before their symptoms become obvious-and possibly at a time when treatment will be most effective. But he speculates babies who go on to become autistic experience the world differently in the first year of life than kids who will not have autism do-and that this altered experience of the world may contribute to subsequent brain development in autism.
Autism Spectrum Disorder (or ASD) is a complex developmental disability characterized by difficulties in social interaction, verbal and nonverbal communication, and repetitive behaviors or interests.
Now autism can reliably be diagnosed as early as age 2, but most kids aren't flagged until after age 4, according to the U.S. Centers for Disease Control and Prevention. In the United States, about one in 68 children has been identified with autism spectrum disorder, according to the Centers for Disease Control and Prevention.More news: Sonos price rises are on the way
Some had older siblings with autism, and were considered high-risk. The researchers took images of the all the babies' brains at six, 12, and 24 months.
Excessive growth in the size and surface area of the brain between six and 12 months of age proved to be a significant predictor, Elison said.
"What we found was that cortical thickness didn't differ between the groups (of infants), but surface area increased at a higher rate than normal between 6 and 12 months of age" in the infants later diagnosed with autism, Piven said, referring to this as "hyper-expansion of cortical surface area".
Research could then begin to examine interventions on children during a period before the syndrome is present and when the brain is most malleable. This brain surface "overgrowth" has been associated with regressed social skills that often crop up in the second year for children with autism.
Numerous measurements the algorithm relied on most are related to surface area, and came from 6-month-old children. Beginning in the 1990s, Piven and other researchers noticed that children with autism tended to have larger brains than developmentally normal children, suggesting that brain growth could be a biomarker for ASD. And the team was pleased with the prediction accuracy.
Still, Mathew Pletcher, vice president and head of genomic discovery at Autism Speaks in New York City, found the research encouraging because "providing early and accurate diagnosis for autism is critical for ensuring the best outcomes".
Despite much research, it has been impossible to identify those at high risk for autism prior to 24 months of age - the earliest time when the behavioral characteristics of ASD can be observed in most children.