Neuromusculoskeletal modelling of myotendinous dynamics during sporting movements of basketball athletes: exploring injury mechanisms

Background: Hamstring and hip adductor muscle strains are the two leading myotendinous injuries sustained by professional basketball athletes. Muscle-tendon unit (MTU) fibre mechanics of hamstring and adductor strain injuries are not well studied, with factors such as fatigue promoted as risk factors in the absence of mechanistic evidence. Invasive direct measurement of MTU mechanics (e.g., muscle fibre length, velocity, and force) is not feasible in the study of healthy populations. Computational musculoskeletal modelling is an alternative approach used in biomechanics research to understand MTU mechanics, examine motor control, simulate surgeries, and for many more applications. This approach can elucidate how an individual’s neural drive to the muscles causes movement, force production, and possibly, injury. Aims/Purpose: The primary aims of this thesis were 1) to streamline and automate time-consuming parts of the neuromusculoskeletal (NMS) modelling workflow, 2) to use electromyography (EMG) in NMS modelling to investigate hamstring and adductor MTU mechanics of basketball athletes, and to elucidate the effect of game-play on MTU mechanics and strain injury risk in the unique high-loading basketball environment, and 3) to compare kinematics, joint moments, MTU mechanics, and the physiological plausibility between modelling methods that use generic and subject-specific bone and muscle geometries. Methods: 1) The performance of five artificial neural networks evaluating sEMG signal quality, a time-consuming aspect of the NMS modelling workflow, were compared. An opensource automated sEMG classification tool was developed and made publicly available. 2) Fibre mechanics of four hamstring (biceps femoris long head, biceps femoris short head, semimembranosus, and semitendinosus) and four adductor (adductor brevis, adductor longus, adductor magnus, and gracilis) muscles during an anticipated cut task performed by ten healthy male basketball athletes pre- and post-game were compared. 3) Muscle fibre mechanics and physiological plausibility of generic and subject-specific NMS models of nine healthy athletes were compared for sprinting and unanticipated sidestep cutting. Additionally, a case study in a single athlete with a unilateral hamstring injury was conducted to illustrate the effects of using generic vs subject-specific NMS modelling in an injured individual. Results: 1) AlexNet demonstrated the highest accuracy (99.55%) with negligible false classifications. The resulting free open software EMG classification tool was provided for any researchers and clinicians using surface EMG. 2) Only biceps femoris long head and semimembranosus muscle forces were significantly lower post-game. No other fibre mechanics were significantly affected by game-play. 3) Compared to generic NMS models, subject-specific models showed significantly lower ankle dorsi/plantar flexion angle during sprinting and several significantly different net joint moments during sprint and cut tasks. Additionally, subject-specific models demonstrated better torque matching, more physiologically plausible fibre lengths, higher fibre velocities, lower muscle forces, and lower simulated activations in a subset of investigated muscles and motor tasks. Conclusions: The conclusions and implications for research resulting from this thesis are threefold: 1) It is possible to streamline and automate sEMG quality evaluation, a time-consuming aspect of the NMS modelling workflow, via ANN without compromising human-like classification accuracy. 2) Mechanistic evidence on hamstring and adductor MTU mechanics during sidestep cutting performed both before and after basketball game-play does not support muscle fatigue as a hamstring and adductor strain injury risk factor in basketball players. 3) Compared to generic models, NMS models that use subject-specific geometry result in more physiologically plausible joint and muscle mechanics and are able to identify between-limb differences. EMG-informed NMS models with subject-specific geometry should be used not only when studying populations with unilateral musculoskeletal pathologies, but also for healthy populations.