【摘要】： 優(yōu)化是放射治療計(jì)劃系統(tǒng)的重要組成部分之一。針對放射治療方案的優(yōu)化問題,本文提出了一系列解決方案,主要包括以下幾個(gè)方面: 1)較全面地介紹了現(xiàn)代最優(yōu)化理論及方法,對NSGA-II多目標(biāo)優(yōu)化遺傳算法進(jìn)行了較深入的研究,提出了在該方法中加入局部搜索算子及外部精英收集池的改進(jìn)算法-NSGA-II多目標(biāo)混合遺傳算法。 2)針對伽瑪?shù)读Ⅲw定向放射治療計(jì)劃系統(tǒng)中的優(yōu)化問題提出了一個(gè)兩步優(yōu)化的方案。首先采用一種改進(jìn)的基于距離變換的幾何優(yōu)化方法,優(yōu)化靶點(diǎn)的數(shù)目、位置及相應(yīng)準(zhǔn)直器直徑;然后根據(jù)幾何優(yōu)化獲得的靶點(diǎn)參數(shù),研究了基于混合遺傳算法GA-BFGS方法優(yōu)化靶點(diǎn)權(quán)重的方法,以避免梯度法可能會陷入局部極小的問題。 3)針對全身伽瑪?shù)吨委燇w部腫瘤的靶點(diǎn)設(shè)計(jì)優(yōu)化問題,提出了一種新穎的靶點(diǎn)設(shè)計(jì)及優(yōu)化解決方案,建立了基于包圍盒的優(yōu)化模型,然后采用遺傳算法進(jìn)行求解,使治療過程更為簡便快捷,并有利于減少治療過程中的定位誤差,填補(bǔ)了全身伽瑪?shù)栋悬c(diǎn)設(shè)計(jì)優(yōu)化研究的空白。 4)針對全身伽瑪?shù)吨委燇w部腫瘤的治療路徑優(yōu)化問題,在靶點(diǎn)優(yōu)化的基礎(chǔ)上,提出了采用遺傳算法進(jìn)行治療路徑優(yōu)化的解決方案,使治療過程更為簡便快捷,有利于減少治療過程中醫(yī)生的勞動強(qiáng)度。當(dāng)靶點(diǎn)數(shù)過多時(shí),我們給出了一種分層優(yōu)化的解決方案,以減少優(yōu)化時(shí)間,滿足臨床的實(shí)際要求。 5)實(shí)現(xiàn)了一種基于NSGA-Ⅱ多目標(biāo)混合遺傳算法的調(diào)強(qiáng)放射治療逆向計(jì)劃的優(yōu)化方法,其中局部搜索采用L-BFGS算法。該方法得到的非劣解在目標(biāo)空間分布均勻,計(jì)算速度快,魯棒性好。 6)在Windows平臺上開發(fā)了一套用于周圍血管內(nèi)放射治療計(jì)劃優(yōu)化的主動漫游虛擬現(xiàn)實(shí)系統(tǒng),用于提取血管的幾何結(jié)構(gòu);建立血管內(nèi)放射治療計(jì)劃的優(yōu)化模型,采用適宜的優(yōu)化方法進(jìn)行優(yōu)化,該方法綜合運(yùn)用了混合遺傳算法GA-BFGS方法和快速模擬退火算法優(yōu)化相關(guān)的參數(shù),其中GA-BFGS算法用于優(yōu)化連續(xù)參數(shù)即放射源的照射時(shí)間,快速模擬退火算法用于優(yōu)化離散參數(shù)即放射源的位置。這一策略綜合考慮了目標(biāo)函數(shù)的形式,BFGS的快速收斂性,遺傳算法的全局收斂性,以及模擬退火算法的全局收斂性。
[Abstract]:Optimization is one of the important components of radiotherapy planning system. Aiming at the optimization of radiotherapy scheme, this paper puts forward a series of solutions, including the following aspects: 1) the modern optimization theory and method are introduced comprehensively. The genetic algorithm of NSGA-II multi-objective optimization is studied deeply. An improved algorithm-NSGA-II hybrid genetic algorithm is proposed to add local search operator and external elite collection pool to the method. 2) for gamma knife stereotactic orientation A two-step optimization scheme for radiotherapy planning system is proposed. Firstly, an improved geometric optimization method based on distance transform is used to optimize the number, position and corresponding collimator diameter of the target, and then according to the geometric optimization of the target parameters, The method of target weight optimization based on hybrid genetic algorithm (GA-BFGS) is studied. In order to avoid the gradient method may fall into the problem of local minimization. 3) for the whole body gamma knife treatment of body tumor target design optimization problem, In this paper, a novel target design and optimization solution is proposed, and an optimization model based on bounding box is established. Then genetic algorithm is used to solve the problem, which makes the treatment process simpler and faster, and helps to reduce the positioning error in the treatment process. Fill in the blank of the whole body gamma knife target design optimization research. 4) aiming at the whole body gamma knife treatment body tumor treatment path optimization problem, on the basis of the target optimization, The genetic algorithm is used to optimize the treatment path, which makes the treatment process simpler and faster, and helps to reduce the labor intensity of doctors in the course of treatment. When there are too many targets, we present a hierarchical optimization solution to reduce the optimization time, 5) an optimization method of intensity modulated radiotherapy reverse planning based on NSGA- 鈪，

本文編號：2226765