Li Dong, Chen Bin, Ran Wei Yu, Wang Guo Xiang, Wu Wen Juan
Xi'an Jiaotong University, State Key Laboratory of Multiphase Flow in Power Engineering, Xi'an 710049, China.
Xi'an Jiaotong University, State Key Laboratory of Multiphase Flow in Power Engineering, Xi'an 710049, ChinabUniversity of Akron, Department of Mechanical Engineering, Akron, Ohio 44325-3903, United States.
J Biomed Opt. 2015;20(9):095014. doi: 10.1117/1.JBO.20.9.095014.
The voxel-based Monte Carlo method (VMC) is now a gold standard in the simulation of light propagation in turbid media. For complex tissue structures, however, the computational cost will be higher when small voxels are used to improve smoothness of tissue interface and a large number of photons are used to obtain accurate results. To reduce computational cost, criteria were proposed to determine the voxel size and photon number in 3-dimensional VMC simulations with acceptable accuracy and computation time. The selection of the voxel size can be expressed as a function of tissue geometry and optical properties. The photon number should be at least 5 times the total voxel number. These criteria are further applied in developing a photon ray splitting scheme of local grid refinement technique to reduce computational cost of a nonuniform tissue structure with significantly varying optical properties. In the proposed technique, a nonuniform refined grid system is used, where fine grids are used for the tissue with high absorption and complex geometry, and coarse grids are used for the other part. In this technique, the total photon number is selected based on the voxel size of the coarse grid. Furthermore, the photon-splitting scheme is developed to satisfy the statistical accuracy requirement for the dense grid area. Result shows that local grid refinement technique photon ray splitting scheme can accelerate the computation by 7.6 times (reduce time consumption from 17.5 to 2.3 h) in the simulation of laser light energy deposition in skin tissue that contains port wine stain lesions.
基于体素的蒙特卡罗方法（VMC）如今是浑浊介质中光传播模拟的金标准。然而，对于复杂的组织结构，当使用小体素来提高组织界面的平滑度且使用大量光子以获得准确结果时，计算成本会更高。为了降低计算成本，人们提出了一些标准，用于在具有可接受精度和计算时间的三维VMC模拟中确定体素大小和光子数量。体素大小的选择可以表示为组织几何形状和光学特性的函数。光子数量应至少是总体素数量的5倍。这些标准进一步应用于开发局部网格细化技术的光子射线分裂方案，以降低具有显著变化光学特性的非均匀组织结构的计算成本。在所提出的技术中，使用了一个非均匀细化网格系统，其中精细网格用于具有高吸收和复杂几何形状的组织，而粗网格用于其他部分。在该技术中，根据粗网格的体素大小选择总光子数量。此外，还开发了光子分裂方案，以满足密集网格区域的统计精度要求。结果表明，在包含葡萄酒色斑病变的皮肤组织激光能量沉积模拟中，局部网格细化技术光子射线分裂方案可将计算速度提高7.6倍（将时间消耗从17.5小时减少到2.3小时）。
