SUMMARY

Work-related injuries due to overexertion remain a leading cause of health problems in manual occupations despite the implementation of safety initiatives. Prolonged exposure to repetitive lifting and twisting under mechanical loading can lead to strain in soft tissues and degradation in bones that can lead to prominent chronic ailments such as low back pain and osteoarthritis which continue to plague the workforce, with about 50% of reported injuries stemming from the back or knee. Chronic bone injuries are difficult to measure in vivo and a considerable amount of deterioration is needed to detect pain in the later stages. Joint contact forces capture the internal force felt by the bone and can be estimated through neuromusculoskeletal modeling; thus, providing insight on internal joint loading. Therefore, there is a critical gap in understanding and mitigating injuries from chronic joint loading in the back and knee.
In the first aim, I seek to develop a framework that characterizes joint contact forces across work-specific lifting tasks in the back and the knee and identifies tasks in need of assistance. Since exoskeletons have shown reductions in energy expenditure, muscle activity, and joint loading, I believe they may be a suitable intervention for harmful joint loading. Consequently, the second aim intends to investigate the effects of exoskeletons on joint contact forces. I hypothesize that prescribing a back and knee exoskeleton can reduce joint contact forces by lowering muscle activity in work-specific tasks. Advances in machine learning have also proven effective at estimating and predicting biological metrics such as kinematics and kinetics from wearable sensors. The objective of the third aim is to create a system to estimate joint contact forces using machine learning technologies and determine the minimal amount of input data required for reliable performance.
Electromyography (EMG) will be used to measure muscle activity and neuromusculoskeletal modeling tools will calculate joint kinematics, kinetics, and contact forces with (1) no exoskeleton, (2) an active knee exoskeleton, and (3) a passive back exoskeleton. Later, a wearable sensor-driven machine learning algorithm will be used to estimate joint contact forces. Completion of the aims will prove beneficial to ergonomists, clinicians, and applied engineers alike by informing rehabilitation strategies, exoskeleton design and controllers, and real-time biofeedback systems.