An Unsupervised Data-Driven Model to Classify Gait Patterns in Children with Cerebral Palsy.

Area of Science:

• Biomechanical analysis
• Pediatric orthopedics
• Data-driven modeling

Background:

• Ankle and foot orthoses are frequently prescribed for children with cerebral palsy (CP).
• Current clinical practice lacks clarity on the reliability of 3D gait analysis (3DGA) for consistent orthotic prescription.
• Data-driven modeling offers potential to uncover complex relationships between 3DGA parameters and orthotic outcomes.

Purpose of the Study:

• To develop a data-driven model for classifying CP gait biomechanics.
• To identify associations between prescribed orthotic types and observed gait patterns.
• To evaluate the utility of 3DGA in guiding orthotic interventions for children with CP.

Main Methods:

• Utilized 3D gait analysis (3DGA) data from typically developing children and children with CP using orthoses.
• Employed unsupervised self-organizing maps and k-means clustering to categorize gait patterns based on gait variable scores (GVSs).
• Analyzed GVSs derived from the gait profile score, measuring deviations from typically developing (TD) gait.

Main Results:

• Identified five distinct pathological gait patterns in children with CP, showing significant differences in GVSs.
• Observed that only 43% of children demonstrated improved gait patterns with orthotic use.
• Found considerable variability in orthotics prescriptions even among children with similar gait patterns.

Conclusions:

• Data-driven modeling can classify CP gait patterns, offering objective insights beyond traditional analysis.
• Current orthotic prescription practices show variability and may not consistently optimize gait outcomes in children with CP.
• Quantitative, data-driven approaches are recommended to enhance the clarity and specificity of orthotic prescription in pediatric CP management.