An automatic approach to assess biomechanical risk using machine learning algorithms and inertial sensors

Area of Science:

• Occupational Health
• Biomedical Engineering
• Artificial Intelligence

Background:

• Work-related musculoskeletal disorders (WMSDs) are a major occupational health concern.
• Manual material handling involves risks like intensity, repetition, and duration.
• Current observational methods for assessing biomechanical risk are subjective.

Purpose of the Study:

• To develop an automated methodology for discriminating biomechanical risk during load lifting.
• To integrate machine learning algorithms with inertial wearable sensors for objective risk assessment.

Main Methods:

• Ten healthy volunteers performed weight-lifting tasks wearing inertial measurement units (IMUs) on the sternum and lumbar regions.
• Inertial signals were processed to extract time-domain and frequency-domain features.
• Machine learning algorithms were applied to classify biomechanical risk levels.

Main Results:

• The system achieved high accuracy (above 86%) and Area Under the Curve (AUC) (above 95%) in discriminating risk classes.
• The sternum provided the most informative data for risk assessment.
• The mean absolute value was identified as the most significant feature.

Conclusions:

• The proposed AI-driven methodology effectively discriminates biomechanical risk associated with lifting.
• Wearable sensors and machine learning offer an objective alternative to subjective ergonomic assessments.
• Further validation in larger populations is recommended for workplace implementation.