Innovative AI-Based Early Detection of Autism Spectrum Disorder Developed

Researchers at the Electronics and Telecommunications Research Institute (ETRI) have introduced a groundbreaking artificial intelligence (AI) system designed for the early screening of autism spectrum disorder (ASD). This advanced technology aims to identify early signs of ASD in young children more efficiently, potentially transforming how early intervention is initiated and improving accessibility for families and healthcare providers.

ETRI's new social interaction recognition AI analyzes infant and toddler behaviors by assessing content designed to evoke social responses within a brief period, typically around six minutes of viewing. Autism spectrum disorder can often be detected through behaviors such as difficulties in social communication, repetitive actions, and limited engagement. Early detection is crucial because it allows for timely interventions that can significantly improve developmental outcomes.

Despite the importance of early screening, current practices face challenges due to a lack of specialists, limited social awareness, and resource constraints, often leading to a delay of two to six years between symptom onset and formal diagnosis. Symptoms can manifest as early as 12 to 24 months, underscoring the need for more accessible diagnostic tools.

Collaborating with Prof. Yoo Hee-Jung from Seoul National University Bundang Hospital’s Department of Psychiatry, the ETRI team analyzed data from over 3,500 cases of children aged 42 months or younger to evaluate the sensitivity and effectiveness of their AI screening method.

Building on this research, the team developed a unique 'social interaction-inducing content' that prompts a variety of social responses like eye contact, pointing, imitation, and response to name, which are then observed and analyzed by AI. This process enables the system to monitor emotional states, gaze behavior, and specific gestures, providing a comprehensive assessment of social communication skills.

The developed technology also employs cameras to observe infants and toddlers in real-time, capturing detailed interaction processes. It can monitor responses such as showing objects, imitating behaviors, and responding to cues, facilitating an objective and quantitative evaluation of social development.

Having established a living lab at the Korea Institute of Robotics and Technology Convergence (KIRO) since 2020, the researchers have been continuously collecting data and testing the AI system on infants and young children, advancing the field of autism screening. This multidisciplinary, convergence-based AI approach addresses the limitations of existing screening tools and offers a promising new method for early diagnosis.

The benefits of this technology extend beyond clinical settings, as it can be integrated into preschools, childcare centers, and even home environments, lowering barriers to mental health services. Its widespread adoption could enhance early detection efforts, promote social awareness of autism, and support preventive testing and intervention strategies.

Dr. Yoo Jang-Hee, lead researcher from the Social Robotics Research Section, highlighted the potential societal impact: "We hope this technology will shorten the time from symptom detection to diagnosis and help change societal perceptions of autism. Our goal is to contribute to solving significant social challenges through innovative research."

Source: https://medicalxpress.com/news/2025-06-ai-technology-early-screening-autism.html