Quantifying Repetitive Hand Flapping Kinematics in Autistic and Non-Autistic Toddlers Using Video-Based Pose...

Area of Science:

• Developmental psychology
• Biomechanical analysis
• Artificial intelligence in healthcare

Background:

• Repetitive behaviors like hand flapping are common in autism spectrum disorder (ASD).
• Distinguishing clinically meaningful differences in hand flapping between autistic and non-autistic toddlers before age 2 is challenging.
• Subtle movement variations may be detectable using advanced technology.

Purpose of the Study:

• To investigate kinematic differences in repetitive hand flapping between autistic and non-autistic toddlers.
• To determine if amplitude and frequency of hand flapping differ between groups.
• To explore the utility of computer vision in identifying early behavioral markers.

Main Methods:

• Utilized AlphaPose, an AI-powered computer vision tool, to analyze video recordings of toddlers during a bubble play activity.
• Quantified the amplitude and frequency of repetitive hand flapping from manually annotated video segments.
• Compared kinematic data between 14 toddlers with ASD and 14 without ASD, aged 13-16 months.

Main Results:

• Event-level analysis revealed significantly greater amplitude of hand flapping in autistic toddlers compared to non-autistic toddlers (p = 0.028).
• No significant differences in the frequency of hand flapping were found between groups (p = 0.464).
• Event-averaged analyses masked these amplitude differences, indicating variability in findings based on analytical approach.

Conclusions:

• Event-level analysis of repetitive hand flapping amplitude may offer a more sensitive method for identifying early kinematic differences in toddlers.
• The findings highlight the importance of fine-grained analysis to understand subtle behavioral variations in early development.
• Computer vision techniques show promise for objective assessment of early developmental behaviors.