A High-Performance Cellular Automaton Model of Tumor Growth with Dynamically Growing Domains
By Jan Poleszczuk and Heiko Enderling
Abstract
Tumor growth from a single transformed cancer cell up to a clinically apparent mass spans many spatial and temporal orders of magnitude. Implementation of cellular automata simulations of such tumor growth can be straightforward but computing performance often counterbalances simplicity. Computationally convenient simulation times can be achieved by choosing appropriate data structures, memory and cell handling as well as domain setup. We propose a cellular automaton model of tumor growth with a domain that expands dynamically as the tumor population increases. We discuss memory access, data structures and implementation techniques that yield high-performance multi-scale Monte Carlo simulations of tumor growth. We discuss tumor properties that favor the proposed high-performance design and present simulation results of the tumor growth model. We estimate to which parameters the model is the most sensitive, and show that tumor volume depends on a number of parameters in a non-monotonic manner.
Keywords: cellular automaton, dynamic boundaries, tumor model, cancer stem cells, sensitivity analysis.
Introduction
Simulating complex multi-scale cellular automata is still a great challenge despite advances in computational power of modern computers in recent years. Cellular automata are increasingly used to simulate tumor growth dynamics...
0 Comments