Computers Detecting Autism in Toddler Talk


A study published yesterday in the Proceedings of the National Academy of Sciences shows that a baby’s “nonsense syllables could contain coded signals” indicating autism, AOL News reported.

A computer program designed by a team led by Kim Oller of the University of Memphis can detect autistic children based on their speech patterns, with an accuracy rate of 85 percent.  Oller hails the voice recognition tool as “the first kind of system that’s totally objective.”  He says the computer system alone isn’t enough to diagnose a child with autism, but that when perfected, it “could be used very effectively” to screen children who seem to be on the spectrum.

The chief scientific officer for Autism Speaks agrees. “It would be very helpful to have an automated way to screen for autism,” said Geraldine Dawson.  She notes, “Not all children on the autism spectrum have language delays,” but “the majority of children with autism do show some delay in the onset of language or early vocalization.”  Diagnosing autism is currently a long and drawn-out process done under the care of a specialist; researchers argue that Oller’s computer program is unbiased and can work much faster than an individual.

“Oller and his colleagues tucked miniature voice recorders into the chest pockets of more than 200 children ages 10 months to 4 years,” according to AOL.  They recorded the children 12 hours a day for approximately one week.  Scientists discovered that autistic children tended to garble their speech “far longer than normal children, making it easy for the software to pick out the autistic kids’ voices.”

This news comes on the heels of the release of the second phase of the Autism Genome Project, showing autistic children carry more copy number variants in their genome than children without autism.

Image: Beverly & Pack via Flickr