A new artificial intelligence (AI) technology has emerged that analyzes infants watching videos to detect autism spectrum disorder (ASD) early. It is expected to make autism diagnosis, which usually takes several years from symptom discovery to actual diagnosis, easier and faster.
The Electronics and Telecommunications Research Institute (ETRI) announced on the 29th that it collaborated with Professor Yoo Hee-jung's team at Bundang Seoul National University Hospital to develop content that induces social interaction and created AI technology to screen for autism by analyzing a 6-minute video of children watching it.
The research team analyzed the sensitivity of autism screening indicators based on 3,531 cases of infants under 42 months. Based on this, they developed an observation scenario for infants that can identify autism using AI technology.
ETRI researchers, in collaboration with Professor Yoo Hee-jung's team from the Department of Psychiatry at Bundang Seoul National University Hospital, analyzed the sensitivity of autism spectrum disorder screening indicators based on 3,531 cases of infants under 42 months.
Based on this, they developed an observation scenario for infants that can be analyzed by AI. The research team produced video content designed to induce social responses from infants, showing interesting subjects, calling their names, and encouraging responses such as imitative behavior, pointing, and eye contact.
The research team captured footage of infants interacting while watching this content and analyzed it with AI. The AI examined whether infants recognized personal traits and emotions, such as their own names, and whether they recognized, imitated, or made eye contact with the actions of others.
Since 2020, the research team has established a living lab for screening autism spectrum disorder at the Seoul Center of the Korea Robotics Convergence Research Institute, continuously conducting observation tests and data collection targeting infants. They have enhanced their technology by gathering data for five years.
The research team expects this technology to play a crucial role in the early diagnosis and intervention of autism. When autism is detected early and appropriate medical intervention is carried out, it can positively impact developmental progress. However, due to a shortage of professionals, limitations in social awareness, and constraints on time and resources, it typically takes 2 to 6 years from symptom discovery to actual diagnosis.
The technology developed by ETRI can be easily used without specialized personnel in nurseries, childcare facilities, developmental centers, and even in general homes. It could serve as a practical solution to activate preventive testing for autism and early intervention, addressing the early screening issues of infants and children. The research leader, Yoo Jang-hee, head researcher at ETRI's Social Robotics Research Lab, said, “I hope this helps shorten the period from symptom discovery to diagnosis along with changes in social awareness regarding autism.”