A research team led by the University of Dubai has developed an artificial intelligence-powered screening tool that could help identify children at higher risk of autism spectrum disorder at an earlier stage, potentially improving access to specialist assessment and early intervention.
The technology combines eye-tracking information with developmental milestone data to produce an early risk indicator for autism. Researchers say the approach could offer healthcare professionals a faster screening option at a time when long diagnostic waiting periods remain a challenge in many healthcare systems.
The new platform also highlights the growing role of behavioral biometrics in healthcare screening. Instead of relying only on questionnaires completed by parents, the system examines measurable patterns in a child’s visual attention, including gaze direction, fixation duration and how the child scans different areas of a screen. These signals are combined with developmental information to provide an additional indication of whether further clinical evaluation may be appropriate.
Autism spectrum disorder can often be identified during the first two years of a child’s life, but many children do not receive a formal diagnosis until considerably later. Delays can limit the time available for families to access developmental support and intervention programmes, which can play an important role in improving outcomes.
The University of Dubai-led research aims to address part of this gap by creating a screening process that is quick, non-invasive and potentially easier to deploy than resource-intensive assessment methods.
AI Screening Combines Eye Movement and Developmental Data
The system uses a tablet or smartphone to monitor how a child looks at short video clips containing social and geometric scenes. While the videos are playing, the technology records eye movements and evaluates patterns in the child’s visual attention.
Parents simultaneously complete a questionnaire covering developmental milestones. Rather than treating the two sources of information separately, the system combines them through an AI model to generate an indicator of autism risk.
This represents a different approach from widely used screening tools such as the Modified Checklist for Autism in Toddlers, or M-CHAT, which relies primarily on information provided by parents or caregivers.
Researchers believe objective eye-tracking data could provide an additional layer of information during the initial screening process. Differences in gaze behaviour and visual attention have been investigated as potential indicators associated with autism, although screening results alone cannot establish a clinical diagnosis.
The technology processes raw eye-tracking measurements, including gaze patterns, fixation duration and visual scanning behaviour. According to the researchers, the system does not first need to convert these measurements into conventional images before analysis.
That design could have practical implications for deployment. By working directly with the underlying eye-tracking information, the system has been developed to operate on relatively accessible devices without requiring high-performance computing infrastructure.
The screening procedure takes approximately two to three minutes, making it potentially suitable for use during routine clinical interactions. Its non-invasive nature could also make the process easier for young children and families compared with more demanding assessment procedures.
Initial benchmark testing produced an accuracy rate of about 96%, which researchers described as an encouraging early result. However, the figure should be interpreted within the context of ongoing research. A promising screening performance does not mean the technology can replace comprehensive developmental assessment or specialist diagnosis.
Faster Screening Could Expand Access to Early Intervention
The potential value of the technology lies not only in its reported accuracy but also in how it could fit into existing healthcare pathways. Autism assessments can require multiple appointments and specialist input, creating delays for families seeking answers about a child’s development.
A rapid screening tool could allow healthcare professionals to identify children who may benefit from further assessment before they enter a longer diagnostic pathway. This could help prioritise specialist resources while giving families an earlier indication that additional evaluation may be necessary.
The researchers also see potential for the technology to move beyond clinical settings. While the current system has been designed for use in healthcare environments, its compatibility with tablets and smartphones could eventually support home-based screening.
Such an expansion could be particularly relevant in areas where specialist developmental services are limited. A digital screening process could provide primary care professionals and families with an additional tool for identifying children who may require specialist attention.
However, broader deployment would require further validation across larger and more diverse populations. Autism presents differently from one child to another, and factors such as age, developmental stage, behaviour and testing environment can influence screening results. Ensuring that the technology performs consistently across different populations will therefore be important before it can become part of routine care.
The development also reflects a wider shift toward using measurable behavioural signals to support healthcare decisions. Eye movement, speech patterns and other forms of observable behaviour are increasingly being investigated as potential sources of information for early screening and assessment.
For the autism field, the immediate opportunity is more targeted: improving the first stage of the diagnostic journey rather than replacing clinicians. If validated in larger studies, the University of Dubai-led system could provide healthcare providers with a rapid way to flag children who warrant closer evaluation.
Early identification remains one of the most important goals in autism care. By combining eye-tracking data with developmental information in a short screening process, the research team is seeking to make that first step more accessible while reducing some of the delays that families currently face.
