Accelerating photon transport in Monte Carlo simulation using octree structure

本研究旨在发展一种利用Monte Carlo模拟的光子/电子输运的高速计算方法。目标描述在有效地执行光子传输中是重要的。为了描述一个对象，已经提出了三种方法：基于形状（SV）的方法，基于体素（VB）的方法，和基于立体几何的方法。然而，这三种方法都有一些缺点（例如，边界精度、计算时间、存储量、手动指令）。针对这一问题，我们采用蒙特卡罗方法，将物体的八叉树表示引入到光子输运中。对象的表示产生每个介质的准确边界和光子传输的快速计算，其精度与使用最精细体素的VB方法相同。我们的Monte Carlo程序首先通过阅读“八叉树串“来建立模拟几何，该”八叉树串“是在模拟前自动从一组物体的连续截面形成八叉树结构而产生的，然后它在几何中传输光子。使用该代码，如果用户只是为他想要模拟光子轨迹的对象准备一组连续切片，则他可以使用由八叉树表示简化的次优几何结构自动执行模拟，而无需通过手写形成最优几何结构。我们的代码的准确性和能力进行了确认，通过比较与电子伽马淋浴代码系统（EGS 4）和模拟使用头部幻影，分别。此外，我们检查了我们的代码使用Zubal的数字幻影（头部和胸部）通过比较结果的VB方法。总之，所提出的方法是有效的任何对象的蒙特卡罗模拟。

The study was carried out to develop a high-speed calculation method in photon/electron transportation using a Monte Carlo simulation. Object description is important in performing the photon transport efficiently. In order to describe an object, three methods have been proposed : a shape-based (SV) approach, a voxel-based (VB) approach, and a solid geometry based approach. However, these three methods have some shortcoming (e.g., accuracy of boundaries, calculation time, amount of memory, manual instructions). To the problems we introduced an octree representation of an object to the photon transport using a Monte Carlo method. The representation of the object yields accurate boundaries of each medium and fast calculation of the photon transport with the same accuracy as the VB method using the finest voxels. Moreover, the octree structure for any object can be formed systematically.Our Monte Carlo code, first, establishes the simulation geometry by reading "octree string, " which is produced by forming an octree structure from a set of serial sections for the object before the simulation automatically, and then it transports photons in the geotry. Using the code, if the user just prepares a set of serial sections for the object in which he wants to simulate photon trajectories, he can perform the simulation automatically using the suboptimal geometry simplified by the octree representation without forming the optimal geometry by handwriting. The accuracy and ability of our code were confirmed by comparison with Electron Gamma Shower code system (EGS4) and simulations using a head phantom, respectively. Moreover, we examined our code using Zubal's digital phantoms (head and thorax) by comparing the results of the VB approach. In summary, the proposed method is efficient in the Monte Carlo simulation of any object.