Local beam angle optimization with linear programming and gradient search
Local beam angle optimization with linear programming and gradient search
复制标题
DOI：
10.1088/0031-9155/52/7/n02
复制
发表时间：
2007-04-07
影响因子：
3.5
通讯作者：
Craft, David
中科院分区：
文献类型：
--
作者：
Craft, David
IMRT 规划中波束角度的优化仍然是一个悬而未决的问题，文献重点关注启发式策略和离散角度网格的穷举搜索。我们展示了如何使用波束角空间中基于梯度的优化以连续方式局部细化波束角集。梯度是使用线性规划对偶理论导出的。将这种局部搜索应用于幻影胰腺病例的 100 个随机初始角度集演示了该方法，并强调了 BAO 问题的多局部最小值方面。由于这种函数结构，我们建议采用彻底的全局搜索的搜索策略，然后在有希望的波束角度集进行局部细化。讨论了非线性 IMRT 公式的扩展。
The optimization of beam angles in IMRT planning is still an open problem, with literature focusing on heuristic strategies and exhaustive searches on discrete angle grids. We show how a beam angle set can be locally refined in a continuous manner using gradient-based optimization in the beam angle space. The gradient is derived using linear programming duality theory. Applying this local search to 100 random initial angle sets of a phantom pancreatic case demonstrates the method, and highlights the many-local- minima aspect of the BAO problem. Due to this function structure, we recommend a search strategy of a thorough global search followed by local refinement at promising beam angle sets. Extensions to nonlinear IMRT formulations are discussed.