Data and model driven multiscale simulation of tumor growth in liver cell, tissue and organ

Since hepatic cancer has a wide spread and is one of the leading causes of death worldwide, computational models become more important for the prediction of the disease and the development of personal treatment. Therefore, the aim of our project is to develop an efficient data-integrated multiscale and multiphase model for tumor growth in the liver. We want to include different size scales, namely the cell, tissue and organ scale, as well as different microscopic substances like treatment agents or nutrient concentrations. Furthermore, the model will integrate the development of fat during a non-alcoholic fatty liver model. The model is based on ordinary differential equations (ODE), describing the metabolic processes on the cell scale, as well as on partial differential equations (PDE) for the description of blood transport and elastic tissue behaviour of the liver lobule. The project aims to enhance this PDE-ODE approach by model order reduction (MOR) via an artificial neural network (ANN) for the cell scale to improve the computational performance of the model. We will train the ANN on experimental and in silico data generated by our project partners so that we get a PDE-ANN approach. To improve the robustness and efficiency of the model, we will develop polymorphic uncertainty quantification procedures via a combination of the variational sensitivity analysis and the Bayesian approach.