Monte Carlo-based Inverse Treatment Plan Optimization for Intensity Modulated Proton Therapy

Abstract

Intensity-modulated proton therapy (IMPT) can achieve a better dose distribution than passive scattering or uniform scanning. For IMPT optimization, Monte Carlo (MC) is desired for spots dose calculations because of the high accuracy in heterogeneous cases. Due to its capability of computing linear energy transfer (LET), MC-based IMPT planning is preferred in biological optimization scheme. However, MC simulation is too slow to be used for this purpose. Although GPU-based MC engine has been developed, the achieved efficiency is still not ideal. The purpose of this work is to develop a new scheme to include GPU-based MC into IMPT. The conventional approach for this purpose is simply using MC repeatedly for each spot dose calculations. However, this is not the optimal approach, because of the unnecessary computations on spots that turned out to have very small weights after IMPT optimization. Memory writing conflict also poses a challenge, if one sequentially compute dose at each spot. To solve these problems, we have developed a new MC-based IMPT plan optimization framework that iteratively performs MC dose calculations and plan optimization. At each dose calculation step, the particles are sampled from different spots based on previously optimized spots intensity map with Metropolis sampling method. Simultaneous handling multiple spots also solves the memory writing conflict problem. We validated the proposed MC-based optimization schemes in one prostate case. It took 5-6 min of total computation time including both spots dose calculation and optimization with only one GPU card for the proposed method, whereas a conventional method naively using MC for spot dose calculations would be  2-3 times slower.