Images revealed which infants would go on to have an autism diagnosis, raising hopes for earlier care and interventions for those affected.
Babies who are most at risk of developing autism as toddlers have been identified from brain scans in the first year of life.
The images helped doctors spot which of a group of children who were already at risk because of autism in the family would later be diagnosed with the condition.
The work raises hopes that affected children could be helped with earlier specialised care and interventions to help their social abilities before any behavioural symptoms start to appear.
Scientists studied 106 babies, all of whom had an older brother or sister with autism, and found that telltale features in brain images taken at six and 12 months revealed which infants would go on to have an autism diagnosis.
The technology is not ready to use in clinics, but researchers who took part in the study believe it lays the groundwork for a tool that can predict autism in high-risk babies before their first birthday.
“We don’t have such a tool yet, but if we did, parents of high-risk infants wouldn’t need to wait for a diagnosis of autism spectrum disorder at two, three or even four years,” said Annette Estes, a co-author on the study and director of the University of Washington in Seattle. “Researchers could start developing interventions to prevent these children from falling behind in social and communication skills,” she added.
About one in 70 children in the UK develops autism, but among those with a diagnosed sibling, the risk of the condition can rise to about one in five. The test has not been tried on children who are not at high risk.
Writing in the journal Nature, the scientists describe how they compared brain scans of high-risk babies with those from 42 low-risk children who had no siblings with autism. On each lab visit, scientists assessed the children’s behaviour and intellectual abilities.
The images, taken as the babies slept, revealed that high-risk infants who developed autism experienced a “hyperexpansion” of the brain’s surface in the first year of life. The unusual growth preceded a rise in brain volume over the following year that was linked to social difficulties the children developed.
The scientists next used a machine learning algorithm to predict from the scans and the infant’s sex which high-risk children would be diagnosed with autism at the age of two. The computer identified eight in 10 of those who developed autism. In about 3% of cases, it predicted autism that did not appear.