ABSTRACT
This chapter addresses the performance requirements for Monte Carlo (MC) simulations to meet the timescale of clinical practice, and we will discuss possibile hardware and software solutions that can speed up MC calculations. The stochastic nature of the interaction is simulated in MC algorithms by sampling distributions obtained from theoretical models and confirmed by experimental data. A large number of particles need to be simulated in order to reduce statistical fluctuations in the scored quantities, e.g. the dose deposited inside the patient. The treatment plan is the set of instructions that a Treatment Planning Software (TPS) produces combining together patient morphology, irradiation directions, dose prescription and accelerator properties. The treatment plan is generated and optimized for each patient by the medical physicists of a treatment center using the TPS. In history-based transport algorithms, as in traditional full-MC codes, each particle is assigned to a GPU core, which follows it step-by-step from particle generation until its removal from simulation.
