An accelerated implementation of Extended Cellular Potts Model for tumor angiogenesis simulations

Modern distributed systems and neural computation face increasing demands in processing complex simulations, particularly in computational biology. While distributed computing offers powerful solutions for large-scale problems, certain computational challenges can be effectively addressed through GPU acceleration alone, providing a foundation for future distributed implementations. In this work, we present a GPU-parallel approach for simulating angiogenesis using the Cellular Potts Model (CPM). Here we deal with how a traditionally sequential biological simulation can be transformed into a parallel implementation, establishing a methodology that could be extended to distributed systems. By implementing the CPM on GPU using CUDA, and incorporating both chemotaxis and electric field effects, we extend the basic angiogenesis model by introducing a tumor-related chemical source, thus making a more detailed and biologically relevant simulation environment, we develop a framework that achieves significant performance improvements while maintaining biological accuracy. Experiments conducted on LEONARDO supercomputer, show up to 39x speedup compared to CPU implementations, particularly in handling large-scale matrices and complex energy calculations.