Artificial intelligence could help experts identify toddlers who may be autistic, researchers said after developing a screening system they say has an accuracy of about 80% for children under the age of two.
The researchers say their approach, which is based on a type of AI called machine learning, could have benefits.
Dr Kristiina Tammimies, a co-author of the study from Karolinska Institutet in Sweden, said: “Using [the] AI model, it may be possible to use available information and identify earlier individuals with increased probability for autism so that they can get earlier diagnosis and help.”
But, she added, “I want to emphasize that the algorithm cannot diagnose autism as it should [still] done with gold standard clinical methods.”
This isn’t the first time researchers have tried to harness AI to screen for autism. Among other studies, scientists have previously used such technology with retinal scans of children.
Writing in the journal Jama Network Open, Tammimies and colleagues report how they harnessed data from a US research initiative called the Spark study which includes information from 15,330 children with a diagnosis of autism and 15,330 without.
The team describes how they focused on 28 measures that can be easily obtained before children are 24 months old, based on parent-reported information from medical and background questionnaires, such as age at first smile.
They then created machine learning models that looked for different patterns in combinations of these features among autistic and non-autistic children.
After using the data to build, tune and test four different models, the team selected the most promising one and tested it on a further dataset of 11,936 participants for whom data on the same characteristics was available. In total, 10,476 of these participants had an autism diagnosis.
The results show that overall the model correctly identified 9,417 (78.9%) participants with or without autism spectrum disorder, with accuracy of 78.5% for children up to two years of age, 84.2% for those aged two up to four years and 79.2% for from four to 10 years old.
A further test with another set of data including 2,854 autistic individuals revealed that the model correctly identified such a diagnosis 68% of the time.
Tammimies said: “This data set was another research cohort with families with only one child with autism and some of the parameters were missing, so the performance was a little lower, which shows that we still need to do more development.”
The researchers said the measures that generally appeared to be most important when it came to the model’s predictions included difficulty eating food, age at first construction of longer sentences, age at achievement of potty training and age at the first smile.
The team added that an additional analysis, comparing participants the model correctly identified as autistic and those incorrectly identified as non-autistic, suggested that the model tended to identify autism in individuals with more severe symptoms and more general development issues.
However, some experts urged caution, noting that the ability for the model to correctly identify non-autistic people was only 80%, meaning that 20% would be incorrectly labeled as possibly autistic. They also noted that pushing for an early diagnosis can be problematic.
Prof Ginny Russell from the University of Exeter said this was because it was difficult to tell which toddlers might have a very severe disability and who would “catch up” despite a slow start.
“My recommendation is below [two years] “It is too early to start applying psychiatric labels based on a few indications such as eating behavior,” she said.
