Combining Postural Sway Parameters and Machine Learning to Assess Biomechanical Risk Associated with Load-Lifting...

Area of Science:

• Occupational Ergonomics
• Biomechanics
• Machine Learning

Background:

• Work-related musculoskeletal disorders are linked to load lifting factors like duration, intensity, and repetition.
• Traditional ergonomic assessments lack standardization, prompting research into new methods.
• Wearable sensors and AI offer a promising avenue for monitoring and mitigating biomechanical risks.

Purpose of the Study:

• To evaluate machine learning models for classifying load lifting risks.
• To use postural sway metrics from an inertial measurement unit (IMU) for risk assessment.
• To align risk classification with the Revised NIOSH Lifting Equation.

Main Methods:

• An IMU placed at the lumbar region captured acceleration data for postural sway analysis.
• Eight participants performed 20 consecutive lifting tasks in two scenarios.
• Eight machine learning classifiers were tested, with Gradient Boost Tree showing the highest performance.

Main Results:

• The Gradient Boost Tree model achieved 91.2% accuracy and a 94.5% Area Under the ROC Curve.
• Feature importance analysis identified key sway parameters and directions influencing risk classification.
• The combination of sway metrics and Gradient Boost model proved feasible for predicting biomechanical risks.

Conclusions:

• The proposed method using IMU-derived sway metrics and machine learning is a viable approach for assessing biomechanical risks in load lifting.
• Further research with diverse participants and lifting conditions is recommended to broaden applicability in occupational settings.