WEARABLE SENSOR-BASED GAIT CLASSIFICATION IN IDIOPATHIC TOE WALKING ADOLESCENTS.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Pediatric Gait Analysis

Background:

• Idiopathic toe walking affects 7-24% of children, increasing fall risk and potential developmental delays.
• Accurate gait classification is crucial for assessing interventions in idiopathic toe walking.
• Wearable sensors and machine learning offer potential for objective gait assessment.

Purpose of the Study:

• To investigate machine learning algorithms for differentiating toe-toe gait from heel-toe gait using wearable sensor data.
• To assess the efficacy of k-means clustering and Long Short-Term Memory (LSTM) models in classifying gait patterns.

Main Methods:

• Five adolescents with idiopathic toe walking wore an inertial sensor at the L5-S1 joint.
• Data from triaxial accelerometers and gyroscopes were analyzed using k-means clustering and LSTM models.
• Feature extraction focused on linear variability metrics like standard deviation and Root Mean Square (RMS).

Main Results:

• K-means clustering successfully differentiated toe walking from typical walking signals.
• Linear variability features were key in distinguishing gait patterns through clustering.
• The k-means model achieved 82% accuracy, 83% specificity, and 86% sensitivity.

Conclusions:

• Machine learning algorithms integrated with wearable sensors can accurately classify idiopathic toe walking gait.
• These techniques hold promise for transforming therapy and monitoring patient progress longitudinally.
• Novel learning-based methods can aid in estimating treatment efficacy for idiopathic toe walking.