PanoVR: A Longitudinal Human-Computer Interaction Study on Promoting Skill Transfer in Children with Autism Spectrum Disorder through Panoramic Multi-sensory Virtual Reality Training

Paper Title

PanoVR: A Longitudinal Human-Computer Interaction Study on Promoting Skill Transfer in Children with Autism Spectrum Disorder through Panoramic Multi-sensory Virtual Reality Training

Publication Info

• Topic area: Virtual reality interventions for Autism Spectrum Disorder (ASD) focusing on skill transfer and neural adaptation.
• Keywords: Autism Spectrum Disorder, panoramic virtual reality, multi-sensory training, skill transfer, effective connectivity, neural adaptation, longitudinal study, human-computer interaction.

Background and Problem

• Problem / challenge: Existing ASD interventions, such as TEACCH, show limited success in improving adaptive behaviors and skill transfer to real-world contexts. Virtual Reality (VR) interventions often lack ecological validity and may induce sensory overload in ASD children.
• Significance: Enhancing skill transfer and adaptability in ASD children is critical for improving their social participation and independence in real-world scenarios.
• Motivation and related work: Prior VR studies have demonstrated potential for skill acquisition in ASD but are limited by low scene realism, lack of interactivity, and insufficient longitudinal validation. Panoramic VR offers high-fidelity environments but has been underexplored for interactive and progressive training in ASD interventions.

Solution

• Proposed approach: PanoVR, a panoramic VR system that integrates progressive multi-sensory stimulation (visual, auditory, olfactory) with staged training from familiar to unfamiliar environments.
• Novelty:
  1. Development of a panoramic VR system with high ecological validity for progressive multi-sensory training.
  2. Integration of behavioral and neural measures (fNIRS) in a longitudinal design to evaluate intervention effectiveness.
  3. Discovery of neural reorganization patterns, including reduced effective connectivity (EC) and between-participant variability, as a mechanism for skill transfer.
• Procedure and key techniques:
  • Panoramic scene construction using real-life settings.
  • Interactive tasks (food allocation and ingredient combination) to enhance motor and cognitive skills.
  • Gradual increase in sensory stimulation (visual, auditory, olfactory) across three training phases: familiar, semi-familiar, and unfamiliar environments.
  • Behavioral and neural data collection using Kinect motion capture and fNIRS.

Results

• Concrete findings:
  • Task completions increased significantly across training stages (T1 to T3), with a 25.5-task improvement from T1 to T3 (p < 0.001).
  • Global EC decreased significantly from T1 to T3 (β = −0.0141, p < 0.001), reflecting more efficient neural coordination.
  • Behavioral metrics (e.g., movement stability, coordination) remained stable across training stages, with 12 out of 15 indicators showing statistical equivalence.
• Advantage over baselines:
  • Enhanced task performance and skill transfer to unfamiliar, high-stimulation environments.
  • Reduced neural redundancy and increased efficiency in cross-regional coordination, particularly between prefrontal and motor cortices.
• Experiments / evaluation:
  • Longitudinal within-subject design with 11 ASD children (ages 6–14) over four weeks.
  • Data collected at three stages (T1: familiar, T2: semi-familiar, T3: unfamiliar environments).
  • Behavioral metrics (e.g., task completion, movement stability) and neural data (fNIRS-based EC) analyzed using linear mixed-effects models and equivalence testing.
• Limitations and future work:
  • Small sample size (n=11) and limited age range (6–14 years).
  • Lack of an untrained control group for comparison.
  • Combined manipulation of scene familiarity and sensory intensity limits disentangling their individual effects.
  • Absence of haptic stimulation and higher temporal resolution in data collection.

Summary

This study introduces PanoVR, a panoramic VR system designed to improve skill transfer and adaptability in children with ASD through progressive multi-sensory training. Over four weeks, participants demonstrated significant improvements in task performance and neural efficiency, with stable behavioral patterns across increasingly complex environments. The findings suggest that controlled sensory progression and stable task structures can support the development of transferable skills and neural reorganization. Future work will address sample size limitations, explore independent effects of training parameters, and incorporate additional sensory modalities to enhance intervention effectiveness.