The Classification of Movement in Infants for the Autonomous Monitoring of Neurological Development

Area of Science:

• Pediatric neurology
• Developmental pediatrics
• Artificial intelligence in healthcare

Background:

• Neurodevelopmental delays following preterm birth or birth asphyxia are common, but diagnosis is often delayed due to subtle early signs.
• Early intervention is crucial for improving outcomes in infants with developmental delays.
• Current diagnostic methods can be inaccessible or require specialized clinical settings.

Purpose of the Study:

• To develop and validate a non-invasive, cost-effective method for automated assessment of infant movements.
• To improve early detection and monitoring of neurodevelopmental disorders in children.
• To enhance accessibility of diagnostic tools through home-based testing.

Main Methods:

• Utilized deep learning and 2D pose estimation algorithms on video recordings of infants (3-12 months) interacting with toys.
• Captured approximately 25 minutes of 2D video data per participant during natural play.
• Classified infant movements based on dexterity and posture during toy interaction.

Main Results:

• Demonstrated the feasibility of capturing and classifying the complexity of infant movements during play.
• Successfully classified infant posture and movement patterns.
• Identified movement features indicative of potential developmental impairments.

Conclusions:

• The proposed method shows promise for assisting practitioners in the timely diagnosis of impaired or delayed movement development.
• Automated analysis of infant movements can facilitate accurate diagnosis and effective treatment monitoring.
• This approach could significantly improve accessibility to early developmental assessments for infants.