June 12, 2017Neurology
Autism affects roughly 1 out of every 68 children in the United States, and siblings of children diagnosed with autism are at higher risk of developing the disorder. Although early diagnosis and intervention can help improve outcomes, there currently is no method to diagnose the disease before children show symptoms. Previous findings suggest that brain-related changes occur in autism before behavioral symptoms emerge.
According to a study published online Science Translational Medicine (7 June 2017), functional connectivity magnetic resonance imaging (fcMRI) may predict which high-risk, 6-month old infants will develop autism spectrum disorder by age 2 years. The study focused on the brain's functional connectivity -- how regions of the brain work together during different tasks and during rest. Using fcMRI, the authors scanned 59 high-risk, 6-month-old infants while they slept naturally. The children were deemed high-risk because they had older siblings with autism. At age 2 years, 11 of the 59 infants in this group were diagnosed with autism.
The authors used a computer-based technology called machine learning, which trains itself to look for differences that can separate the neuroimaging results into two groups -- autism or non-autism -- and predict future diagnoses. One analysis predicted each infant's future diagnosis by using the other 58 infants' data to train the computer program. This method identified 82% of the infants who would go on to have autism (9 out of 11), and it correctly identified all of the infants who did not develop autism. In another analysis that tested how well the results could apply to other cases, the computer program predicted diagnoses for groups of 10 infants, at an accuracy rate of 93%.
Overall, the team found 974 functional connections in the brains of 6-month-olds that were associated with autism-related behaviors. The authors proposed that a single neuroimaging scan may accurately predict autism among high-risk infants, but caution that the findings need to be replicated in a larger group.