The tool, known as CAMI, short for Computerized Assessment of Motor Imitation, uses a standard video in which a child is asked to mimic the movements of a person on screen.
The AI system then evaluates how the child imitates those simple movements, a behavioral marker linked to autism.
In a clinical study, researchers screened 183 children between the ages of 6 and 13. According to the researchers, the AI tool achieved an accuracy rate of 80% to 85% in diagnosis.
Kaleab Kinfu, a Ph.D. candidate involved in the research, said the technology is designed to focus on a child's imitation behavior rather than environmental factors captured in the video.
"There are nuisance factors in this video where the camera was positioned, lighting conditions. Our AI is trained to ignore all these and focus on the quality of the imitation, and then produces a score, and we use that score to see if that child is in the spectrum or not."
Traditional autism evaluations rely on behavioral observations, developmental histories and standardized questionnaires completed by parents about their child.
Researchers said the goal of CAMI is not to replace physicians, but to provide clinics, schools and families with a faster diagnostic tool.
Professor Rene Vidal said the technology offers a more detailed assessment of motor behavior than current standardized testing methods.
"This is a much more fine-grained motor assessment than is usually done in the current standardized test," Vidal said.
He went on to say that researchers envision the tool becoming part of a broader diagnostic process.
"And so the idea is that this could become an automated component of the complete diagnosis," he said.
Researchers said the next phase of the project will focus on expanding testing to a larger population, including adults, and improving CAMI's accuracy rate.