Data-driven modeling of avascular tumors

Keywords: Approximate Bayesian inference, PDE models, Convex optimization, Data mining, Cell-based models

Aims: The purpose of this project is to investigate the feasibility of strict data-driven modeling of tumors, specifically avascular tumors, that is, before any blood vessels have been recruited. Starting from a basic PDE-model of reaction-diffusion character in two space dimensions, we will develop methods to examine how well given data fits the model and how to best select the model parameters. Such techniques may be relied upon to assess the future development of the tumor as well as its response to treatment with drugs.

In the second part of the project we will confront the previously developed PDE-model with actual data from our collaborators. The grand challenge is of course to be able to parameterize this data in a truthful way and to accurately predict its dynamics. Even a moderate success in this part opens the way for extremely interesting machine learning tasks as well as for advances of useful in silico modeling.

Supervisors: Stefan Engblom and Robin Eriksson