Korean researchers develop social interaction cognitive AI technology for early screening of autism spectrum disorder
Apr 29, 2025
|
The Electronics and Telecommunications Research Institute (ETRI) announced that it has developed a 'social interaction-inducing content' for infants and children and a 'social interaction-cognitive AI' technology that can select autism spectrum disorders by analyzing video images within 6 minutes collected while watching them.
The research was jointly conducted by Professor Yoo Hee-jung of Bundang Seoul National University Hospital, Professor Kim Hong-guk of Gwangju Institute of Science and Technology (GIST), and Kim Min-kyu, director of the Korea Institute of Robot Convergence.
Autism spectrum disorders, which can be detected by observing certain behaviors and developments, such as lack of social communication skills and limited and repetitive behaviors, can have a positive effect on the progress of development if early detection and appropriate medical interventions are made, so early screening and intervention are very important. However, it is reported that it takes two to six years from symptom detection to actual diagnosis due to lack of professional manpower, limitations in social awareness, and time and resources.
Major symptoms of autism spectrum disorder can appear 12 to 24 months of age and even before that in some cases, further emphasizing the importance of early screening and intervention.
The ETRI researchers collaborated with Professor Yoo Hee-jung of the Department of Mental Health at Seoul National University Bundang Hospital to analyze the sensitivity of the autism spectrum disorder screening index based on data from 3,531 infants under 42 months old and developed an infant observation scenario that can be applied with artificial intelligence technology.
Based on this scenario, we have developed the world's first 'social interaction-inducing content' that can induce and observe various social reactions such as ▲response when showing an interesting object or ▲ calling a name ▲ imitation behavior ▲ pointing ▲ eye contact. In addition, we developed a 'social interaction recognition AI" technology that photographs the interaction process of infants and toddlers watching the content with a camera and performs ▲ personal characteristic information and emotion recognition ▲ detection of gaze and call response ▲ gesture recognition ▲ detection of imitation and homogeneous behavior.
Dr. Jang-Hee Yoo of ETRI's Social Robotics Laboratory, Director of Research, stated "We hope that it will help shorten the period from symptom detection to diagnosis along with the change in social perception of autism."
This article was translated by Naver AI translator.
