A 20-minute brain scan may one day be able to detect autism in children, a new study suggests.

A group of international researchers, including University of Southern Queensland’s Professor Raj Gururajan, are developing an artificial intelligence (AI) system for diagnosing Autism Spectrum Disorder (ASD) from brain signals and neural patterns.

In a major breakthrough, the researchers were able to create a unique classification of ASD from Electroencephalogram (EEG) recordings of children’s brain activity.

More than 150,000 people in Australia are estimated to have autism, with the disorder affecting about one in 160 children.

Professor Gururajan, a researcher in information systems with expertise in health informatics, said early diagnosis was important for early intervention and ensuring good outcomes.

“This system could improve patient outcomes by a great deal because the earlier we’re able to detect and diagnose children with autism, the sooner families can start to make decisions about therapies, treatments and supports,” he said.

Published in the journal Complex & Intelligent Systems, the study examined and tested 18 nonlinear, highly distinctive features extracted from two sets of EEG signals from children aged four to 13.

One set belonged to 40 children with a diagnosis of ASD. The other belonged to 37 children with no diagnosis of neuro developmental disorder.

The researchers identified three distinct combinations of features that were strongly predictive of ASD.

The next stage of the research will focus on strengthening the model with more data and testing, with the aim of developing a cloud-based system that can automatically detect ASD in a single brain scan.

Professor Gururajan said the system would help take the guesswork out of autism diagnosis.

“Currently, diagnosing ASD can be very difficult because there is no medical test or blood test that can be used to get a definitive diagnosis. Instead, doctors look at the child’s behaviour and development over a period of time, sometimes years, to make a diagnosis,” Professor Gururajan said.

“There is a real need for an automated diagnostic tool to aid healthcare professionals in diagnosing ASD accurately and early, particularly clinicians who may not have a lot of experience with children with autism.”

The proposed system will work by feeding live data from a brain scan into a cloud service that uses machine learning algorithms to analyse and rapidly classify the EEG signals into two categories: ASD and non-ASD.

“Storing the data in cloud means it could also be used to diagnose other neurological diseases that affect children, such as attention-deficit hyperactivity disorder and epilepsy,” Professor Gururajan added.