Mechanical modelling and imaging of human tendons

Tendons are fibrous connective tissues which are saturated with fluid. Cells within the tendon control the production and organisation of tendon constituents in response to local mechanical signals. Tendinopathy is a painful condition characterised by collagen disruption, accumulation of proteoglycans (protein-sugar complexes), and altered cell behaviour. Exercise is both a key aetiological factor for tendinopathy and a means for rehabilitation. High-frequency high-magnitude forms of loading can be detrimental to tendon structure and cause pain, especially when performed with small rest intervals between loading bouts. In contrast, low-frequency high-magnitude forms of loading have been demonstrated to reduce pain in tendinopathy and increase tendon stiffness. Evidence suggests that the disruption to tendon structure seen in tendinopathy is due to a maladaptive cellular response to loading. However, to date there is no clear understanding of how the spatially heterogeneous mechanical environment within the tendon subject to load, and the resulting changes in tendon material properties coordinated by cellular processes, contributes to damage or rehabilitation. Evidence to date suggests that tensile strains promote type I collagen synthesis and cause an increase in tendon stiffness, while high fluid flux or shear may lead to a reduction in collagen content and increase in proteoglycans, leading to a reduction in the stiffness and increase in permeability. In this thesis, we aim to combine physical models of tendon deformation with imaging techniques to identify why high- and low-frequency loading regimes have differential effects on tendon stiffness and permeability. Using a poroelastic adaptation model for tendon we investigate how the local mechanical environment contributes to beneficial rehabilitative and maladaptive changes in tendon material properties in response to loading. The spatial distribution of tendinopathic changes predicted by the adaptation model are then contrasted with imaging results generated by a machine learning approach which identifies tendinopathic regions, providing mechanistic insight into why damage occurs in certain regions. We first derive a poroelastic model for tendon to identify differences in the local mechanical environment when the tendon is subject to cyclic tensile high- and low-frequency loading. The results are used to infer which local mechanical stimuli promote a maladaptive response by the cells and which promote an adaptive response. Increases in proteoglycans and reductions in collagen content have been posited to cause a reduction in the permeability and stiffness, respectively. Hence, we also investigate how localised reductions in permeability and stiffness alter the local mechanical environment, to identify whether changes in collagen and proteoglycan content amplify or reduce the mechanical stimuli that occur with high- and low-frequency loading. We proceed by combining the poroelastic tendon model with a set of phenomenological rules prescribing the changes in tendon stiffness and permeability, which serve as surrogates for changes in type I collagen and proteoglycans, in response to local mechanical stimuli occurring during loading. The phenomenological component is time-dependent, allowing us to incorporate the effects of recovery time between loading bouts. A series of loading bouts are simulated and stiffness and permeability changes predicted by the model are contrasted with the global stiffness measurements and qualitative changes in the proteoglycans observed clinically. To complement the preceding work, we develop an automated means for the texture-based segmentation of ultrasound images of the tendon. We utilise a hidden Gaussian Markov random field (HGMRF) model to automatically segment a dataset of tendinopathic and healthy ultrasound images of the Achilles tendon. The algorithm provides a means to identify and quantify disruption to healthy collagen architecture on imaging, allowing us to gain greater insight into the spatial pattern of changes that occur in tendinopathy. Using the poroelastic model we show that high fluid fluxes occur with high-frequency loading while the axial strain remains comparable between different loading frequencies. A set of phenomenological rules in which reductions in the permeability and stiffness occur in response to high fluid flux, and increases in stiffness occur in response to axial strain, are able to predict clinically observed material property changes seen with high- and low-frequency loading scenarios. The model predicts tendinopathic changes to occur near the transverse tendon border due to high fluid flux, which is consistent with the location of damage predicted by the HGMRF-based segmentation approach.