Evaluating Wearable Sensor Technologies for Predicting Shoulder Endurance

This study investigated three distinct methods, advancing from non-wearable to wearable technologies, to predict endurance times (ETs) for dynamic shoulder flexion/extension tasks in conjunction with a torque-based biomechanical endurance model to aid in the prevention of work-related musculoskeletal disorders. The study investigated if the use of different wearable sensors affected endurance predictions and if a fully wearable approach is feasible in the context of real-time worker monitoring in conjunction with this model which uses a new approach to generate maximum torque inputs. Our findings indicated that the integration of wearable sensors significantly affected ET predictions and presents potential for the development of a wearable-based system with real-time, predictive fatigue capabilities integrating our endurance modelling work. Notably, the use of inertial measurement units to derive joint angles produced the most precise endurance predictions, with an absolute mean error of 24.8%. Conversely, predictions based on motion capture data exhibited a higher absolute mean error of 30.2%. When pressure insole data were used to estimate dumbbell mass, the absolute mean error was 29.8%, though this approach often underestimated ETs due to overestimating the dumbbell mass. Future work will focus on enhancing the accuracy of these predictions in real-time scenarios and real work environments.Clinical Relevance-This establishes the potential of wearable sensor integration with biomechanical endurance models to enhance real-time fatigue prediction, aiding in the prevention of work-related musculoskeletal disorders.