Abstract:
We present a novel parallel 3-D model of tumor proliferation. To simulate the growth dynamics of normal, cancerous and vascular tissues we use a hybrid method integrating cellular automata (CA) with N-body off-grid paradigm, so called, complex automata model (CxA). The interacting particles with dynamically evolving attributes may represent a single cell (cancerous or normal) or a fragment of blood vessel. Therefore, to mimic realistic tumor masses, huge ensembles of particles have to be simulated on multi-core processor systems. There exist many methods widely used for parallelization of classical N-body dynamics. However, they cannot be applied directly in our CxA model, because the evolution of tumor system is controlled by additional processes such as: cell life-cycle, nutrients and TAFs (tumor angiogenesis factors) diffusion and blood flow. These processes influence physical states of particles, e. g., their type, size and ability for proliferation/annihilation. We show that despite these difficulties, particle model can be efficiently implemented on small multi-core processor systems achieving almost linear speedup for as many as 8 threads and speedup of about 30 on 64 threads UltraSPARC T2 processor. We show that this model framework allows for simulating up to 1 million particles in a reasonable time using modest computer resources.

Paper available at ACM.