Machine Learning Based Abnormal Gait Classification with IMU Considering Joint Impairment.

Area of Science:

• Biomechanics
• Rehabilitation Engineering
• Medical Technology

Background:

• Gait analysis is crucial for evaluating motor function in rehabilitation and elderly care.
• Existing systems face challenges in accurately identifying specific joint impairments.

Purpose of the Study:

• To develop and optimize an abnormal gait classification algorithm using inertial measurement units (IMUs) and walkway systems.
• To differentiate between normal and impaired gaits, and to identify specific joint disorders (knee and ankle).

Main Methods:

• Ten healthy participants simulated normal, knee-impaired, and ankle-impaired gaits under varying bracing conditions.
• Feature extraction utilized Recursive Feature Elimination with Cross-Validation (RFECV).
• Classification models were built using Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGB).

Main Results:

• The IMU-based system achieved over 91% accuracy in classifying three gait types.
• The walkway system achieved less than 77% accuracy, struggling to distinguish between knee and ankle impairments.
• IMU data provided superior discrimination for gait abnormalities.

Conclusions:

• The IMU-based system demonstrates significant potential for accurate gait assessment in individuals with joint impairments.
• This technology could enhance rehabilitation strategies and patient management.
• Further research is recommended for clinical application refinement.