Virtual Reality Based Adaptive Game for Enhancing Joint Attention in Children With Autism

Area of Science:

• Neuroscience
• Developmental Psychology
• Computer Science

Background:

• Autism spectrum disorder (ASD) is characterized by atypical brain development, often manifesting as deficits in joint attention, a key social-communication skill.
• Children with ASD exhibit gaze behaviors that impede coordinated attention, impacting social interaction and learning.
• Existing virtual reality (VR) systems for ASD interventions have limitations in adaptability and rely heavily on performance-based learning.

Purpose of the Study:

• To propose and evaluate a novel virtual reality and adaptive game-based system (VRAGS) for assessing and enhancing attention in children with ASD.
• To develop an advanced AI model for accurate prediction of attention levels in children with ASD.
• To provide a customized and adaptable tool for therapeutic intervention and improved learning capacity in children with ASD.

Main Methods:

• A VRAGS framework was developed, involving data acquisition within a 3D game environment.
• Data preprocessing utilized improved tanh normalization, followed by feature extraction including raw, statistical, and improved entropy features.
• Attention prediction was performed using a hybrid classification (HC) model integrating an Adam-optimized deep convolutional neural network (AODCNN) and a bidirectional long short-term memory (Bi-LSTM) network.

Main Results:

• The proposed HC model achieved high performance metrics: 0.933 accuracy, 0.925 F-measure, and 0.936 negative predictive value (NPV).
• The AODCNN model incorporated improved tanh activation and Adam optimization for stable convergence and robust feature learning.
• Experimental evaluations demonstrated that the VRAGS framework significantly outperformed existing methods in attention prediction.

Conclusions:

• The developed VRAGS framework shows significant potential as a customized and adaptable tool for therapeutic intervention in children with ASD.
• The hybrid AI model effectively captures spatial patterns and temporal dependencies for precise attention level prediction.
• This approach supports improved learning capacity and offers a promising avenue for addressing attention deficits in ASD.