Development of a Parallel Electron and Photon Transport Code (PMCEPT) I: Method and Absorbed Dose Computation in Water

The Monte Carlo (MC) method for high-energy photon and charged particle transport is the most accurate for dose calculations in radiotherapy. However, the large amount of computing time required by general purpose MC codes has prevented their use for routine dose distribution calculations for customized radiation treatment planning. One of the best ways to provide an accurate dose distribution within an acceptable time limit is to develop a parallel MC simulation algorithm on a beowulf PC cluster. We developed a parallel MC electron and photon transport simulation code (PMCEPT) based on the standard message passing interface (MPI). Our MC results agreed with those of the MCNP5 code developed by the Los Alamos National Laboratory and widely used in many hospitals. By preventing the superposition of a series of random numbers on different processors, our parallel results agreed well with the serial ones. The parallel efficiency approached 100 %, as was expected. A feasibility study for building a customized cancer radiation treatment planning simulation system showed that a Linux beowulf PC cluster with 35 up-to-date processors is adequate.