Overview
A research team from the Human-Tech Institute at Universitat Politècnica de València has created a new system aimed at the early detection of Autism Spectrum Disorder (ASD) utilizing virtual reality and artificial intelligence. This system boasts an accuracy rate exceeding 85%, significantly improving upon traditional detection methods that typically rely on manual psychological assessments.
Research Findings
- The study was published in the journal Expert Systems with Applications.
- Researchers analyzed children’s movements while they engaged in various tasks within a virtual reality setting to identify effective AI techniques for detecting ASD.
Benefits of Virtual Reality
According to Mariano Alcañiz, director of the Human-Tech Institute, the use of virtual reality provides:
- Realistic environments that elicit authentic responses from children, mimicking their daily interactions.
- A more genuine understanding of autism symptoms compared to traditional laboratory tests.
System Functionality
The virtual system operates by projecting a simulated environment onto a large screen or the walls of a room, capturing the child’s movements through a camera as they perform various tasks. This method:
- Standardizes autism detection by analyzing biomarkers related to behavior, motor activity, and gaze direction.
- Requires only a large screen and an affordable camera, making it accessible for early intervention settings.
Advancements in AI
Researcher Alberto Altozano, who developed the AI model with Professor Javier Marín, noted that:
- The new deep learning model outperforms traditional AI techniques in identifying ASD.
- The system processes children’s movements during the virtual experience to generate a diagnosis that enhances both accuracy and efficiency.
Long-Term Research Collaboration
This innovative system is the culmination of eight years of research in collaboration with the Red Cenit cognitive development center. Recent doctoral research by Eleonora Minissi validated the system’s effectiveness and highlighted the importance of motor activity as a promising biomarker for autism detection.
Future Implications
The findings suggest that the AI model can be adapted for analyzing motor behaviors in various tasks, paving the way for further exploration into the motor characteristics of autistic children during activities such as walking or talking.
For more information, refer to the study: Altozano, A., et al. (2025). Introducing 3DCNN ResNets for ASD full-body kinematic assessment: A comparison with hand-crafted features. Expert Systems with Applications. doi: 10.1016/j.eswa.2024.126295