中文摘要：

目的：调强放射治疗是治疗癌症的一种主要手段。对放射治疗计划进行优化是实现调强放射治疗的关键步骤。传统的直接子野优化方法比较耗时,为了改进此缺点,本文提出一种基于共轭梯度法的快速直接子野优化方法。方法：将多叶光栅的叶片位置和子野的权重同时纳入问题模型,将该直接子野优化问题建模为一个非线性数学优化问题。直接求解该问题比较困难,采用＂分而治之＂的策略,将其分解为两个较容易的子问题。第一步采用共轭梯度法来优化子野权重,第二步通过移动叶片的方法来优化子野形状,两个步骤交替进行,最终在较短时间内找到满足要求的治疗方案。结果：本文对两个临床病例：胰腺肿瘤病例和头部肿瘤进行了测试。对胰腺肿瘤病例的优化用了3 min 20 s,优化后靶区和危及器官的剂量体积直方图以及等剂量线均满足临床要求。对头部肿瘤病例的优化用了9 min 37 s,尽管该病例包含较多的勾画器官,器官分布较集中,对优化提出了较大挑战,但优化后靶区和危及器官的剂量体积直方图以及等剂量线还是能较好的满足临床要求。结论：本文提出的方法很好的解决了调强放射治疗中一个复杂的非线性优化问题,对临床两例实际病例的测试结果显示,该方法可以在较短时间内找到满足要求的治疗方案,有较好的临床实用价值。

英文摘要：

Objective Intensity Modulated Radiation Therapy （IMRT） is a main method to cure cancer and optimizing the treat-ment plans of radiation therapy is a key step to implement radiation therapy. The traditional direct aperture optimization methodis time-consuming. To overcome this, a fast conjugate gradient-based direct aperture optimization method was proposed. Meth-ods The leaf positions of multi-collimator and weights of apertures were both included in the problem model. And it was formu-lated as a nonlinear mathematic optimization problem. To solve it directly was difficult, divide-and-conquer strategy was adopt-ed and the problem was divided into two easier problems. First, conjugate gradient algorithm was used to optimize the weight ofapertures. Second, optimizing the shapes of apertures by moving the MLC leaves. The combination of the two steps could give agood resolution in a relatively short time. Results Two clinical cases： pancreas cancer and head cancer were tested.Optimizationof the pancreas cancer took 3 minutes 20 seconds. Optimization results of Dose-Volume histogram and isodose line showed thatthe treatment plan could meet the conventional clinical requirement. Optimization of the head cancer took 9minutes 37 seconds.Though this case includes more delineated organs that distribute close to each other, which was a big challenge to optimization,optimization results of the Dose-Volume histogram and isodose line showed that the treatment plan could still meet the conven-tional clinical requirement well. Conclusion The proposed method could solve a ditficult nonlinear optimization problem in IM-RT well. Test results of two clinical cases showed that the method could generate a resolution meeting the clinical requirementin a relatively short time, and has a good practical value in clinic.