• Posted November 21, 2023

AI Might Accurately Spot Autism in Early Childhood

University of Louisville researchers say they've developed an artificial intelligence (AI) system with a near-perfect record of diagnosing autism in toddlers.

Using specialized MRI scans of the brain, the tool diagnosed toddlers with 98.5% accuracy, according to findings scheduled for presentation in Chicago next week at a meeting of the Radiological Society of North America (RSNA).

Research presented at meetings is typically considered preliminary until published in a peer-reviewed journal.

"Our algorithm is trained to identify areas of deviation to diagnose whether someone is autistic or neurotypical," Mohamed Khudri, a visiting research scholar who was part of the team that developed the system. said in an RSNA news release.

The AI system relies on DT-MRI, a special technique that detects how water travels through the brain along what are known as "white matter tracks."

The AI system isolates images from DT-MRI scans and and looks at markers revealing the level of connectivity between brain areas. A machine learning algorithm compares patterns in the brains of children with autism to those of normally developed brains.

"Autism is primarily a disease of improper connections within the brain," co-author Dr. Gregory Barnes, a professor of neurology and director of the Norton Children's Autism Center in Louisville, said in the news release. "DT-MRI captures these abnormal connections that lead to the symptoms that children with autism often have, such as impaired social communication and repetitive behaviors."

Researchers tested their method with 226 children between 24 and 48 months of age from the Autism Brain Imaging Data Exchange-II. Of those, 100 were developing normally; 126 were affected by autism.

The AI approach was 97% accurate in spotting real cases of autism (avoiding false-positive reports) and it was 98% accurate in identifying children who did not have autism. It's overall accuracy was 98.5%, the team concluded.

"Our approach is a novel advancement that enables early detection of autism in infants under two years of age," Khudri said. "We believe that therapeutic intervention before the age of three can lead to better outcomes, including the potential for individuals with autism to achieve greater independence and higher IQs."

Fewer than half of children with autism received a developmental evaluation by 3 years of age, according to new report from the U.S. Centers for Disease Control and Prevention, and 30% of those who meet criteria for autism were not diagnosed by age 8.

Researchers cited several reasons for delayed diagnosis, including lack of resources at testing centers. Khudri said the AI system could help speed the process.

It produces a report detailing which brain pathways are affected, the likely impact on function and a severity score that can be used to guide intervention.

"The idea behind early intervention is to take advantage of brain plasticity, or the ability of the brain to normalize function with therapy," Barnes said.

Researchers are seeking clearance for the AI software from the U.S. Food and Drug Administration.

More information

Autism Speaks has highlights of the 2023 report on autism from the U.S. Centers for Disease Control and Prevention.

SOURCE: Radiological Society of North America, news release, Nov. 21, 2023