SU-E-T-523: Runtime Optimization for the Automatic Monte Carlo Dose Computation System MC2

Abstract

Purpose: Optimizing runtime of a user friendly automated Monte Carlo dose computation system for proton treatments using passively scattered and intensity modulated proton therapy plans. Methods: The MC2 code is a dose computation system based on the Monte Carlo system MCNPX, which is routinely used at the MD Anderson Cancer Center. Besides the successful implementation and use of the code, the efficiency can be optimized to minimize runtime on a cluster of LINUX servers. Based on the 3‐dimensional statistical uncertainty values scored during the simulation of several test proton beams in water, we have developed an optimization procedure for the minimal number of source particles required to achieve desired statistical uncertainty inside the target volume for arbitrary patient fields. Results: First Monte Carlo dose calculations of proton doses in patients were provided in the timeframe of one to several days. Optimization of the initial number of source particles for each proton energy (in IMPT) or range modulator wheel rotation (in PSPT) resulted in significant decrease in computation time. Conclusion: Current efforts focus on optimization of runtime for patient specific proton plans. Among many aspects, which include variance reduction techniques, such as using increased cell importance inside the CT volume, weight window applications and others, the focus of this effort was the minimization of the number of histories to achieve acceptable statistical uncertainties.