Monitoring muscle activity in pediatric SCI: Insights from sensorized rocking chairs and machine-learning

Area of Science:

• Pediatric Rehabilitation
• Biomedical Engineering
• Rehabilitation Robotics

Background:

• Activity-based therapy is crucial for enhancing trunk control in pediatric spinal cord injury (SCI).
• A novel sensorized rocking chair prototype was developed to engage trunk muscles.
• This research focuses on predicting muscle activation using data from this sensorized chair.

Purpose of the Study:

• To utilize sensor data from a rocking chair to predict muscle activation in children.
• To evaluate the efficacy of different machine learning models in predicting muscle use.
• To assess the correlation between sensor-derived features and actual muscle activity.

Main Methods:

• Collected sensor data (forces, accelerations, movement) and electromyography from children with SCI and typically developing children (2-12 years).
• Employed multiple linear regression, decision tree regression, and neural network models to predict muscle activation.
• Performed correlation analysis to determine individual sensor contributions to prediction accuracy.

Main Results:

• Neural network models demonstrated superior performance compared to regression-based models.
• Multiple linear regression and decision tree regression showed significant correlations for a subset of children with SCI.
• Neural network predictions achieved significant correlations for all participating children with SCI.

Conclusions:

• Embedded sensors in the rocking chair effectively capture data relevant to muscle activation.
• Machine learning techniques, particularly neural networks, offer a promising approach for predicting muscle activity during therapy.
• Further research is needed to refine prediction models and validate generalizability for clinical application.