ABSTRACT
Monte Carlo (MC) simulation is considered as the most accurate method for radiotherapy dose calculations due to its capability of faithfully describing the physical processes and flexibly handling complicated geometries. Since the MC simulation is a stochastic method, a large number of particle histories are simulated to achieve a desired statistical accuracy. Despite the vast advancement in computer architecture and the increase of processor clock speed in recent years, the efficiency of the currently available MC dose engines is still not completely satisfactory for routine clinical applications in radiotherapy. One straightforward way to mitigate this issue is to perform the computation in a parallel fashion by taking advantages of advanced parallel computer architectures. By distributing the total computation load to available computing units, it is conceivable that a significant speedup factor can be achieved. Over the years, there has been a considerable amount of research regarding how to implement various MC dose calculation packages on a variety of computing architectures, including multicore central processing unit (CPU), CPU cluster, graphics processing unit (GPU), cloud computing, and so on. The use of those parallel processing techniques for MC dose calculation has offered an attractive approach toward fast MC dose calculation in clinically realistic environments.
