【作者】 李永杰；

【导师】 尧德中；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2004， 博士

【摘要】 放射治疗与手术治疗和化学药物治疗一起,组成了肿瘤的三大治疗手段。从经典的三维适形放射治疗技术发展到现在的调强放射治疗技术(intensity-modulated radiotherapy, IMRT),是放射肿瘤学史上的一次重大变革。但鉴于实际问题的复杂性,IMRT的优势还远没有在临床应用中完全发挥出来。IMRT治疗计划作为调强放疗的核心和基础,其中尚有许多急需解决的问题。本论文算法研究和临床需求紧密结合,在实现和完善常规的IMRT优化技术的基础上,围绕射野方向优化、提高优化速度、缩短治疗时间等方面的问题进行了深入、细致和创造性地研究。所做的主要工作如下:(1)综述了国内外IMRT的研究现状;分析了目前IMRT优化算法的不足和临床需求;总结了共轭梯度法(conjugate gradient, CG)和遗传算法(genetic algorithm, GA)的特点和性能;阐述了剂量计算模型。这些工作为后续研究奠定了基础。(2)以物理目标函数为基础,实现了基于共轭梯度法的射野强度分布的优化。其中为了提高优化的性能,我们采用了独特的辅助器官、动态调节剂量点权重和剂量特征矩阵索引等技术。实例研究表明,该算法具有速度快、效果好、用户输入参数少等优点,已在临床应用中取得了很好的效果。(3)针对目前常用的静态IMRT治疗中存在的子野个数多、治疗时间长等不足,首次提出了用GA+CG的方法来优化子野,达到了实用水平。该技术充分结合了GA方法的优点和实际问题的特点,创造性地设计了一套GA的编码和遗传操作模式,并在优化过程中同时考虑了多叶准直器(multileaf collimator, MLC)的物理限制。该算法被国外论文评阅人评价为在解决子野优化问题上的首创。(4)针对目前方法在射野方向优化时所需时间太长的不足,本文创造性地提出并实现了基于“GA+CG”的射野方向优化技术。该技术用GA和CG来分别优化射野的方向和射野强度分布。设计了针对具体问题的编码方式、遗传操作和独特的免疫操作处理。通过事先计算并保存能量特征矩阵和适应度标定等技术,保证了优化的高效性。(5)首次提出并初步建立了基于专家知识的自动选择射野方向的技术(KBASE)的基本框架和流程。KBASE技术充分把临床专家的丰富经验与优化算法相结合。在优化中成功地利用了射野方向约束和计划模板两种类型的专家知识。初步的研究结果表明,KBASE是可行和有效的。(6)对本论文的优化技术的临床实施进行了剂量验证。在两种MLC上的初步测试结果表明,本工作提出的IMRT的剂量计算和实施过程是准确可靠的。更多还原

【Abstract】 Radiotherapy, together with the surgery and chemotherapy, are the three main means for tumor treatment. It is a historic advancement for tumor treatment that the classical three-dimensional (3D) conformal radiotherapy (3DCRT) evolved into the intensity-modulated radiotherapy (IMRT). Whereas, the advantages of IMRT have not yet been fully utilized, because of the complicated clinical conditions. The IMRT planning, one of the key issues of IMRT application, still has many problems needed to be dealt with.Aiming at the clinical requirements, this thesis realizes the conventional IMRT optimization technique, i.e. the optimization of the beam intensity maps, which is the base of the other researching works. Then researches are done on the beam angle optimization, on shortening optimization and saving treatment time, and so on, which are briefly listed below:(1) A comprehensive review is made on the current status of the IMRT, including the shortcomings of the current IMRT optimization algorithms and the clinical requirements. Also, a summarization is done to the algorithms used in this thesis: conjugate gradient method (CG) and genetic algorithm (GA), and a brief description is given to the photon dose calculation algorithm. These review and summarization are essential preparation for the coming researches.(2) Based on a physical objective function, a method for the optimization of beam intensity maps is developed using CG, among which some special steps are taken in order to improve the performance of the optimization. The results of simulated and clinical tumor cases demonstrate that the proposed algorithm is valid and timesaving. The algorithm has been embedded into a commercial TPS system for routine IMRT planning. (3) Aiming to reduce the number of the segments and shorten the treatment time for static IMRT (step-&-shoot), the technique of GA-based deliverable segment optimization (GADSO) is originated by the author. In GADSO, some novel genetic operations are designed, and also the MLC physical constraints are incorporated. GADSO has been published on Phys. Med. Biol., and is appraised as the first work to introduce GA into the segment optimization.(4) The author creatively develops the automatic beam angle selection (ABAS) using the hybrid algorithm of GA and CG, in order to overcome the extensive computation of the currently available optimization methods. In ABAS, the beam angles and the intensity maps are treated as two separated groups of variables and optimized using GA and CG, respectively. The efficiency of ABAS is achieved by taking some special measures, such as problem-dependant genetic operations, immunity process, fitness scaling, and so on. Some clinical applications show that ABAS is valid and feasible. ABAS has been accepted as a paper by Phys. Med. Biol..(5) As the first one, the author proposes and develops a framework for the knowledge-based beam angle optimization (KBASE), in which the plentiful clinical experiences accumulated over time by the physicists oncologists are incorporated into <WP=7>the GA optimization procedure. In KBASE, two types of expert knowledge are utilized: (a) beam orientation constraints, which are used to reduce the angle searching space, and (b) template plans, which are mainly used to guide the GA genetic progress. Some preliminary results validate the presented KBASE.(6) As a supplementary work, the clinical dose verification is made for Vairan MLC and Siemens MLC, and the comparison results are shown for the point doses and film dose distributions, among which good agreement is achieved.更多还原