Study of Calculation Method for 3D Dose Distribution and Range Distribution in a Patient Body in Proton Therapy

利用实际质子治疗所获得的CT数据，对质子辐照病人进行了蒙特卡罗模拟。从计算时间、质子辐照二次粒子效应模型以及利用蒙特卡罗模拟产生的剂量分布函数等方面考察了该模拟方法在质子治疗计划系统中的实际应用能力。将GEANT软件应用于质子治疗中的模拟。这是一个基于粒子的完全模拟代码，包括核相互作用。患者结构是由体素构成的，体素的属性由CT数据决定。在质子疗法中，模拟需要每个体素中材料的原子序数Z和原子质量A。因此，在质子模拟中，CT数据中的电子密度应转换为Z和A。有效原子序数和原子质量定义为：Zⁱⁱ<Σω</sup>，Aⁱ，Efⁱ</sup>=Σω_iA_i，ω_i=P_iA_i/Σ^IA，其中…更多的e P_i是材料中第i个组分的数权重。使用从模体数据确定的校准曲线将它们表示为CT数据的函数。每层5 mm厚的CT数据有320×320个像素，大小为1 mm×1 mm。患者结构被构建为一组体素，如在治疗计划系统中所计划的在模拟代码中设置的。在蒙特卡罗模拟中，我们建立了如下三点模型：(1)计算一次通过每个体素的质子的能量沉积和运动学参数值；(2)在补偿器(团注过滤器)的前端制作只有每个脊滤器和每个产生的质子能量的空间和运动学分布表；(3)使用扩展的深度剂量分布来估计初级和次级质子通过体素的能量沉积。在用GEANT程序进行蒙特卡罗模拟时，在以下计算条件下，产生患者体内剂量分布的计算时间为1-2小时：一台PC工作站(Linux OS，Pentium4)，1个切片(5 mm×2 mm)的每个体素截面的1%精度，在250 MeV垂直于质子束半径为7 cm的圆内。蒙特卡罗模拟中每个事件计算的独立性使得使用n台并行的PC工作站可以减少1/n的计算时间。这些PC工作站集的集群化使统计事件更容易进行n次汇总。我们认为，该方法可以作为质子治疗中的治疗计划系统，具有多台并行的PC工作站。较少

Monte Carlo simulation of proton irradiation into a patient has been performed using CT data obtained for real proton treatment. We examined practical capability of this simulation method in treatment planning system in proton therapy in a point of calculation time, a model of secondary particle effects in proton irradiation, and possibility in use of dose distribution function produced with the Monte Carlo simulation. The GEANT is applied to the simulation in the proton therapy. It is a particle-based full simulation code including nuclear interaction. A patient structure is constructed with voxels, whose properties are determined by CT data. In the proton therapy, the simulation requires an atomic number Z and an atomic mass A of the material in each voxel. Thus an electron density in the CT data should be converted into Z and A in the proton simulation. Effective atomic number Z_<eff> and atomic mass A_<eff> are defined as : Z_<eff>=Σω_iZ_i, A_<eff>=Σω_iA_i, ω_i=P_iA_i/ΣP_iA_i, wher … More e P_i is a number weight of the i'th component in a material. They are expressed as a function of the CT data using calibration curves determined from phantom data. Each 5mm-thick slice of the CT data has 320×320 pixels in 1mm×1mm size. The patient structure is constructed as a group of voxels, as setup in the simulation code as planned in the treatment planning system. We developed the model with the following three points in the Monte Carlo simulation : (1) calculating energy deposit and kinematical parameter values once for a proton passing through each voxel, (2) making tables of spatial and kinematical distribution of only protons for each ridge filter and each proton energy in production at the front of compensator (bolus filter), and (3) using extended-depth-dose distribution in estimating energy deposits of primary and secondary protons passing through voxels. In introducing this model to the Monte Carlo simulation with the GEANT code, calculation time is 1-2 hours in producing dose distribution in a patient body under the following calculation condition : one PC workstation (Linux OS, Pentium4), 1% accuracy for each voxel cross-section of 1 slice (5mm×2mm), in a circle of 7cm in radius perpendicular to the proton beam at 250MeV. Independence of each event calculation in the Monte Carlo simulation makes it possible to reduce computing time 1/n with using n parallel sets of PC workstation. Clusterization of these PC workstation sets makes statistics n times easily in summing up events. We consider that this method is capable as treatment planning system in proton therapy, with several parallel sets of PC workstation. Less