Technological breakthrough in early autism detection

The LabLENI from the Human-Tech Institute (UPV), in collaboration with the Red Cenit cognitive development center, has developed an innovative early detection system for Autism Spectrum Disorder (ASD) in children aged 3 to 7. This pioneering tool combines virtual reality (VR) and artificial intelligence (AI), achieving diagnostic accuracy rates above 85%.
The system non-invasively analyzes children's movements, gaze direction, and behavior as they interact with various virtual environments. These scenarios are projected onto walls or large-format screens, integrating the child’s own image and tracking their reactions via a camera. This method provides a more natural, affordable, and accessible alternative to traditional psychological tests, while also offering objective data through behavioral and motor-based biomarkers.

“Virtual reality allows us to simulate everyday situations that provoke authentic responses—far more realistic than those typically seen in laboratory settings,” explains Mariano Alcañiz, director of the Human-Tech Institute.
The AI model was developed by researchers Alberto Altozano and Javier Marín, who compared traditional techniques with a deep learning approach. The results, published in the journal Expert Systems with Applications, show that the new model achieves higher accuracy and can identify ASD across a broader range of tasks within the VR experience.
This research is part of the European project T-EYE, focused on developing new technological tools for the early detection of neurodevelopmental disorders. As part of the project, researcher Eleonora Minissi recently presented her doctoral thesis, in which she validated the VR-based system and analyzed the effectiveness of different biomarkers. Her findings highlight the diagnostic value of atypical motor patterns—an area of ASD that has so far received limited attention.

“The ease of capturing these motor data, along with their strong diagnostic potential, make motor activity a highly promising biomarker for autism,” says Minissi.
The Human-Tech team has spent over eight years working in collaboration with Red Cenit to design, validate, and refine technological tools for early ASD detection. The researchers note that the AI model could be adapted to analyze other motor behaviors—such as walking or conversational gestures—opening new avenues for exploring the motor characteristics of autism.