Apparent damage accumulation in cancellous bone using neural networks

In this paper, a neural network model is developed to simulate the accumulation of apparent fatigue damage of 3D trabecular bone architecture at a given bone site during cyclic loading. The method is based on five steps: (i) performing suitable numerical experiments to simulate fatigue accumulation of a 3D micro-CT trabecular bone samples taken from proximal femur for different combinations of loading conditions; (ii) averaging the sample outputs in terms of apparent damage at whole specimen level based on local tissue damage; (iii) preparation of a proper set of corresponding input–output data to train the network to identify apparent damage evolution; (iv) training the neural network based on the results of step (iii); (v) application of the neural network as a tool to estimate rapidly the apparent damage evolution at a given bone site. The proposed NN model can be incorporated into finite element codes to perform fatigue damage simulation at continuum level including some morphological factors and some bone material properties. The proposed neural network based multiscale approach is the first model, to the author’s knowledge, that incorporates both finite element analysis and neural network computation to rapidly simulate multilevel fatigue of bone. This is beneficial to develop enhanced finite element models to investigate the role of damage accumulation on bone damage repair during remodelling.