Part of a collaboration between various universities from U.S.A and Canada, a recent study provides evidence for the potential of using brain-imaging methods and computer (deep-learning) algorithms to identify autism in children that are at high familial risk (that is those having an older sibling with autism), as early as their first year of life.
Two groups of infants: 106 at high-risk (15 high-risk that met the autism criteria; 91 high-risk that did not met the autism criteria) and 41 low-risk (had no sibling with autism) participated in the study with evaluation taking place at 6, 12 and 24 months of age. The evaluations included behavioural measures of autism symptoms (ADOS and CSBS scores) and Magnetic Resonance Imaging (MRI) of the brain (measures such as total brain volume and surface area).
The researchers found significant differences between groups* in the total brain volume at 24 months but not at 6 or 12 months. These findings confirm previous observations suggesting that children with autism show increased brain volume and head circumference as early as two years. Also, significant differences in the surface area growth from 6 to 12 months were found between the groups* with those at high-risk that met the autism criteria showing higher rates.
In addition, a significant relationship was found between the surface area growth rate and the brain overgrowth for all subjects. Even more, the researchers identified that the rate of brain volume overgrowth (12-24 months) is linked to autism severity (as measured by ADOS social affect score, at 24 months).
Computer (deep-learning) algorithms were used to investigate whether and which brain-imaging measures at 6 – 12 months of age could be used to accurately identify those infants who would later meet the criteria for autism at 24 months of age. It was found that, among infants with an older autism sibling, the brain-imaging measures at 6 and 12 months of age successfully identified 80% of the children who would be clinically diagnosed with autism at 24 months. Also, analysis of these algorithms suggest that surface area information (at 6 and 12 months of age) is a primary predictor for the autism diagnosis.
The results of this study may have important implications for early detection and intervention for autism. However, further work is needed in order for the computer algorithms to be used as a potential clinical tool for identifying autism in high-risk infants. Also, it should be noted that even though a proof of principle, the findings of this study should not be applied to the larger population of children with autism who do not have an older sibling with autism.
For a more detailed description of this study see the full article here.
*the differences were between the group of high-risk children that met the autism criteria and the low-risk as well as high-risk children that did not met the autism criteria.