Infant movement classification through pressure distribution analysis.

Area of Science:

• Biomedical Engineering
• Developmental Neuroscience
• Clinical Diagnostics

Background:

• Neuromotor disorders, such as cerebral palsy, require objective early detection methods.
• Differentiating typical infant general movements, specifically fidgety vs. writhing movements, is crucial for developmental assessment.
• Current assessment methods may lack objectivity and scalability for widespread clinical use.

Purpose of the Study:

• To introduce and evaluate a non-intrusive pressure sensing device for classifying infant general movements.
• To differentiate between fidgety movements (fidgety period) and writhing movements (pre-fidgety period) in infants.
• To assess the potential of this approach for early detection of neuromotor disorders.

Main Methods:

• A cohort of 45 typically developing infants was studied from 4 to 16 weeks post-term.
• Multi-modal sensor data, including pressure data from a 1024-sensor mat, was collected across seven laboratory sessions.
• 1776 five-second pressure data snippets were analyzed using various neural network architectures (SVM, FFNN, CNN, LSTM) for movement classification.

Main Results:

• The convolutional neural network (CNN) achieved the highest classification accuracy at 81.4%.
• The pressure sensing approach demonstrated high potential for clinical applications when compared to other automated assessment methods.
• The study successfully distinguished between fidgety present and absent movement patterns.

Conclusions:

• The pressure sensing approach facilitates efficient, large-scale motion data acquisition and sharing.
• This technology has the potential to be scalable for daily clinical application in evaluating infant neuromotor functions.
• Further improvements to the approach could enhance its utility for early diagnosis and monitoring of developmental disorders.