Master of Science, The Ohio State University, 2022, Industrial and Systems Engineering
While biomechanical models can be insightful to potential mechanisms that can contribute to low back pain, they require significant computing power and various technologies that are not readily available in settings such as clinics. On the other hand, lightweight sensing systems such as inertial measurement unit sensors (IMUs) can be integrated in wearable technologies to easily capture high level spine kinematics. While kinematics can be useful for identifying pain populations or tracking motion metrics for patients, they do not provide a look at the internal structures of the spine during dynamic motions. Several studies have attempted to bridge this disparity. Specifically, muscle coactivity of the trunk muscle groups can greatly impact the overall loading and shear forces on the spine, and several studies have attempted to only use kinematics to predict the trunk muscle forces during static exertions. The goal of this study was to present a methodology that can predict trunk muscle forces as a time series, during dynamic motions, which does not require electromyographic (EMG) signals. This was achieved by first, collecting a new dataset with motions that capture key spine kinematics using an EMG-assisted biomechanical model. Second, to use kinematic-derived variables from this data with deep learning methodologies to predict the trunk muscle forces.
30 healthy subjects performed a series of unloaded, bending, and twisting, during a standardized spine motion assessment while wearing EMG sensors on ten trunk muscles. Several variables were extracted from the time series, including velocities, muscle lengths, height, weight, and torso angles. Using several deep learning architectures, these variables were trained to map these variables to the ten muscle forces produced during the dynamic motions.
Various deep learning architectures were investigated, however only three main architectures were reported and pursued. Assessed on an independent test set, the architec (open full item for complete abstract)
Committee: William S. Marras (Advisor); Samantha Krening (Committee Member)
Subjects: Industrial Engineering