Development of the p-medicine oncosimulator as a parallel treatment support system

Abstract

The purpose of this research was the parallelization of the Wilms' Oncosimulator, an integrated cancer treatment support system modeling the growth of nephroblastoma tumors and their in vivo response to chemotherapeutic modalities. In this concept, the Oncosimulator has been optimized in order to perform efficient computations on the newest heterogeneous parallel architectures: the CPU and GPU based computing architectures. The simulator has been implemented using a novel solution for distributed computing on heterogeneous architectures - the Cactus computational toolkit with CaKernel as the module for the computations performed on computing accelerators. In this publication the challenges faced during the process of porting the Oncosimulator onto the aforementioned architectures (within CaKernel framework) are addressed, and the performance benefits of such approaches are analyzed. The successful parallelization of the Oncosimulator advances its computational efficiency and enhances its reusability as well as its eventual translation into clinical practice. The research was performed in the context of the p-medicine project*.