Autism Spectrum Disorder (ASD) can be a difficult condition to deal with. It can affect person’s social and communication skills, behavior and emotional state. The good news is that the sooner the treatment begins, the better are the outcomes. Now, a team of researchers from the University of North Carolina, Chapel Hill, has developed a deep learning algorithm that can predict with high accuracy whether a young child is at high risk of developing ASD.
The new algorithm can detect brain growth changes linked to autism in children as young as 6 months old. Its high accuracy makes it a valuable addition to ASD testings, since most current behavioral questionnaires only have a 50% accuracy.
In the study, the team enrolled 106 infants with an older sibling with an autism diagnosis, and 42 infants with no family history of autism. All of the children’s brains were scanned – 6 months old, 12 months old and 24 months old – and while the researchers saw no change in any of the children’s overall brain growth between 6 and 12 month mark, they did see a significant increase in the brain surface area of the high-risk children who were later diagnosed with ASD.
Since previous studies have shown brain volume enlargement in ASD, this study simply added to that insight by pinpointing that this change (brain enlargement) happens between 12 and 24 months. The new study also showed that during the first year, surface area enlarges.
Using the information about brain volume and surface area, as well as the fact that boys are more likely to develop ASD, the algorithm was able to identify eight out of ten children with the disorder.
The success of the new algorithm will reflect in earlier diagnoses and intervention, which in turn will reflect in better outcomes for people with ASD.
References:
IEEE Spectrum (http://spectrum.ieee.org/the-human-os/biomedical/imaging/ai-predicts-autism-from-infant-brain-scans)
Digital Trends (http://www.digitaltrends.com/cool-tech/autism-algorithm-81-percent/)
The Wonderful World of Algorithms 