Virtual Reality Helps Detect Autism Earlier and More Accurately Than Traditional Methods

A High-Accuracy System

A team from the Human-Tech Institute at the Universitat Politècnica de València (UPV) has introduced a new system for early detection of autism spectrum disorders (ASD), combining virtual reality and artificial intelligence. Their approach achieves an accuracy rate of over 85%, marking a significant improvement over traditional methods that often rely on subjective assessments by psychologists or parents. The results of this research were published in the journal Expert Systems with Applications.

Natural Reactions Are Key

The aim of the study was to verify whether virtual reality can contribute to more accurate autism diagnosis. During the research, children immersed themselves in a virtual world and completed simple tasks without feeling pressured. Meanwhile, a camera discreetly recorded their movements, facial expressions, and gaze direction—data that might easily be missed in standard testing. The AI then processed the data, searching for patterns indicative of an autism diagnosis.

The findings show that the AI-based approach is not only more efficient but also, in many ways, more accurate than commonly used diagnostic tools. According to the researchers, the naturalness of the reactions is crucial: virtual environments mimic real-life conditions, allowing children to behave more spontaneously than during traditional testing.

Another advantage is the method’s technical simplicity: the system only requires standard equipment (a camera and a large screen), making it an accessible and cost-effective solution. Accurate early diagnosis may thus become a (not just virtual) reality for more children than ever before.

Motor Skills as a Key Diagnostic Aspect

The AI model was developed by researcher Alberto Altozano in collaboration with Professor Javier Marín. Their goal was to create a system capable of detecting ASD across a broader range of tasks and with higher precision than previous methods.

"The designed model can identify ASD more accurately and across more types of interactions within a virtual environment," said Altozano. The algorithm is capable of processing data in real time, significantly speeding up the entire diagnostic process.

Researcher Eleonora Minissi also contributed to the study by testing the system in her doctoral thesis and analyzing various biomarkers. Her findings showed that motor activity—a frequently overlooked area in diagnostics—can play a key role in detecting autism.

"Motor skills are easy to measure and highly promising for early diagnosis. Yet we still underestimate them and pay less attention compared to, for example, social or visual aspects of ASD," Minissi noted.

A Path Toward Deeper Understanding

The UPV team spent eight years developing the system in collaboration with the Red Cenit cognitive development center, where testing was conducted with children diagnosed with ASD. The latest results show that AI can be further developed for diagnostic use—for example, in analyzing gait or verbal communication.

"This opens the door to a deeper understanding of the motor symptomatology of autism," added Mariano Alcañiz, director of the Human-Tech Institute. Autism diagnosis may thus become faster, more objective, and more accessible. If such technologies expand beyond research settings, they could help detect ASD symptoms early in thousands of children who might otherwise go undiagnosed and unsupported.

Editorial Team, Medscope.pro

Sources:
1. Altozano A., Minissi M. E., Alcañiz M., Marín‑Morales J. Introducing 3DCNN ResNets for ASD full‑body kinematic assessment: A comparison with hand‑crafted features. Expert Systems with Applications, 2025; Vol. 270: 126295, doi:10.1016/j.eswa.2024.126295.

2. Revolutionary System Enables Early Detection of Autism Spectrum Disorder. Bioengineer.org, April 7, 2025. Available at: https://bioengineer.org/revolutionary-system-enables-early-detection-of-autism-spectrum-disorder