Dubai researchers develop AI tool to support earlier autism screening in children

Researchers in Dubai have developed an artificial intelligence-powered screening tool that combines eye-tracking technology and developmental information to identify children who may benefit from specialist autism assessment.

A research team led by the University of Dubai has developed an artificial intelligence (AI) tool designed to support earlier screening for autism spectrum disorder in children.

The technology analyses children’s eye movements alongside information about developmental milestones to generate an early risk indicator.

Researchers hope the system could help identify children who may require further assessment sooner, potentially reducing delays in accessing specialist services.

During the screening process, children watch short videos containing social and geometric scenes while eye-tracking technology records how they look at the material.

The system analyses information including gaze patterns, fixation duration and visual scanning behaviour.

At the same time, parents complete a developmental milestone questionnaire.

The AI model then combines both sets of information to produce an early risk indicator that can help determine whether a child should be prioritised for specialist assessment.

Unlike some conventional AI systems, the technology analyses the raw eye-tracking data directly rather than first converting it into images.

Researchers said this approach allows the software to operate more efficiently on tablets and smartphones without requiring high-performance computing equipment.

The screening process takes around two to three minutes and is non-invasive.

Initial benchmark testing returned accuracy of around 96 per cent, which researchers described as promising.

However, the research team stressed that the technology is designed to support early screening and referral rather than provide a formal diagnosis.

Clinical validation studies are continuing across different healthcare settings.

The researchers ultimately hope the technology could be used in clinics and, potentially, in homes to broaden access to early screening.

Autism can often be identified during the first two years of life, while many children are not formally diagnosed until around four years of age, according to the Gulf News report.

The researchers said earlier identification could help children access appropriate intervention and support sooner.

The new system differs from screening approaches that rely predominantly on parent-completed questionnaires, such as the Modified Checklist for Autism in Toddlers.

By combining developmental information with objectively measured eye movements, the researchers are investigating whether AI can provide an additional source of information to support screening decisions.

The project is funded through the Dubai Research, Development and Innovation Grant Initiative, established within the Dubai Future Foundation.

It is led by the University of Dubai’s College of Engineering and Information Technology in collaboration with Emirates Health Services and Al Amal Psychiatric Hospital.

Researchers from three universities with Australian connections are also involved – the University of New South Wales, Macquarie University and the University of Wollongong in Dubai.

While autism screening and diagnosis sit within health and specialist assessment pathways, developments in early identification are relevant to early childhood education and care (ECEC).

Educators regularly observe children’s communication, interactions, play, learning and development over time and may contribute observations as part of conversations with families and other professionals.

The emergence of AI-assisted screening tools therefore raises broader questions about how technology could complement existing developmental monitoring and referral pathways.

The Dubai research is still undergoing clinical validation, meaning its effectiveness across different settings and populations will need to be established before broader implementation.

The researchers have also been clear that the system is intended to assist screening and prioritise specialist referrals, not replace clinical assessment or provide an autism diagnosis.

Read  the research here.