Using Bayesian data selection to improve modelling of biological tissue mechanics with application to tendons

Using Bayesian data selection to improve modelling of biological tissue mechanics with application to tendons #

Jessica E Forsyth, James Casey, Tom Shearer, Simon Cotter

15:30 Monday in 3Q68.

Part of the Cell and tissue mechanics session.

Abstract #

Tendons and ligaments are crucial for stabilisation and movement of joints within the body. They are predominantly made of collagen and have a complex, hierarchical microstructure. Experimentally quantifying the parameters defining the microstructure is non-trivial and there is often a high level of uncertainty due to experimental errors. Instead of measuring these parameters directly, we can use microstructural mathematical models to infer their values from macroscale stress-strain data. The experimental stress-strain data used to fit the models often includes data where damage has already started to occur within the tendon which results in inaccuracies of the model fits and thus the inferred model parameters. Within this work we use the novel approach of hierarchical Bayesian data selection - to infer on not only the model parameters, but also what stress-strain data should be used to fit the model to. By incorporating Bayesian data selection into our model, we are better able to determine at what strain damage within the tendon, or deviation from the chosen model, starts to occur and thus obtain better estimates of the tendon microstructure parameters.