图像引导放疗中区域化剂量验证方法的研究

Current practice of radiation therapy (RT) involves optimization of treatment plans using physical criteria (i.e. maximization of target dose and minimization of doses to critical structures), so dosmetric verification is very critical before RT. It is composed of dose calculation and dose delivery, and the corresponding verification results are generally evaluated by γ index and dose-volume histogram (DVH). The γ index is introduced to delineate the accuracy of dose deposition, but it is just a statistical value without the regional information on target volume (TV) and organs at risk (OAR). Although DVH is calculated based on the dose distribution of TV and OAR, it is always calculated by simplified algorithm or single density phantom to improve computational speed, leading to inaccuracy and anatomical information loss. In this work, we will use Alderson Rando anthropomorphic phantom to simulate real anatomical structures. A series of CT scans will be performed with different field-of-view (FOV) settings and slice thickness to evaluate the impact of volume pixel on γ index and DVH. Patients with cancers such as nasopharyngeal carcinoma and lung cancer will be chosen. After a RT plan is designed, we will create the corresponding verification plan through a commercial treatment planning system (TPS) with special algorithm, for example pencil beam convolution (PBC), analytical anisotropic algorithm (AAA), Acuros XB, and XVMC. At the same time, we will re-calculate the dose distribution of the above verification plan using Monte Carlo (MC) simulation. Comparing the difference between MC and the commercial TPS, we will discuss the accuracy of γ index and DVH in different regions. When the verification plan is delivered, a large number of factors will influence the results of RT, for example the rotational degrees of gantry and collimator, and the positions of jaw and multi-leaf collimator (MLC). Herein, we will reconstruct the verification plan with random errors and systematic errors using MC method based on the impact of the above four factors. Next, we will measure the actual dose distribution with dosimeters such as RT film, and obtain the impacts of gantry, collimator, jaw and MLC on γ index and DVH. In general, after a verification plan is performed, the verification uncertainties are composed of errors in dose calculation and those in dose delivery, but we can't get the delivery uncertainty directly. In our analyses, we have got the accuracy of dose calculation at the beginning, so we can obtain the delivery contribution by deducting the dose-calculation uncertainty out of total uncertainties. After this work, we should re-schematize quality assurance (QA) of patient RT, dose calculation verification plus routine RT QA. It is different from the conventional QA, dosmetric verification plus routine QA. In the new QA frame, we can improve RT efficiency due to the omitted overlapping jobs because dose delivery verification is a part of routine RT QA.

由于放疗是以肿瘤致死高剂量、正常组织辐照低剂量为基本出发点，所以剂量验证（剂量计算和剂量传输）格外重要，其结果常用γ指数和剂量体积直方图（Dose-Volume Histogram, DVH）评估。γ指数以单一数值表示剂量沉积准确度，不提供靶区、危及器官等区域性信息；DVH虽含区域信息，却多用简化算法、单一密度模体，造成准确度下降、解剖信息丢失。本课题采用Rando模拟真实解剖结构，获取不同扫描野（Field Of View, FOV）及层厚CT影像，评估体素对γ指数和DVH影响；以肺癌、鼻咽癌等作为研究对象，运用Pencil Beam Convolution (PBC)、Analytical Anisotropic Algorithm (AAA)等算法制订Rando模体验证计划，对比Monte Carlo（MC）结果，分区域讨论γ指数和DVH计算准确性；分析机架、机头、铅门、多叶准直器（Multi-Leaf Collimator, MLC）执行位置偏差，利用MC重构具有特定误差的Rando模体计划，借助剂量仪测定并计算上述四因素对各区域γ指数和DVH影响；统计剂量计算和剂量传输贡献。

剂量验证（剂量计算和剂量传输）是放疗过程极其重要的一个环节，其结果常用γ指数和剂量体积直方图（DVH）评估。γ指数是以单一数值表示剂量沉积准确度，不提供靶区、危及器官等区域性信息；DVH虽含区域信息，却多用简化算法或单一密度模体，造成准确度下降或解剖信息丢失。.本课题获取了不同扫描视野（FOV）、不同层厚条件下的非均匀密度仿真人体以及25例肺癌患者的CT影像，评估体素对γ指数和DVH的影响。研究发现像素尺寸和层厚对体积计算的精度均有影响，且层厚影响最大。像素尺寸对剂量学参数的影响无统计学差异；随CT层厚的增加，靶区适形指数（CI）减小，脊髓最小剂量（Dmin）、平均剂量（Dmean）和心脏接受30 Gy照射剂量体积占比（V30）、40 Gy 剂量体积占比（V40）增大，而肺DVH则不受影响。.以35例鼻咽癌为研究对象制定MapCheck放疗验证计划，运用各向异性分析算法（AAA）计算剂量分布，与Monte Carlo（MC）计算结果进行对比，研究发现两种算法的γ通过率无明显差异；以吻合距离（DTA）3 mm和剂量偏差（DD）3%为标准，MC算法的通过率为（99.5±0.3）%，AAA算法的通过率为（99.6±0.2）%，p>0.05，无统计学差异。以25例肺癌患者为研究对象，分别运用两种算法计算剂量，对比发现靶区内95%的体积所接受的剂量（D95%）的MC计算值要低6.5%；而全肺的低剂量区域（≤10 Gy）MC计算平均值要高于AAA算法。.模拟机架、机头、铅门、多叶准直器（MLC）的位置偏差发现<1°的机架、机头角度误差所引起的靶区及危及器官的各剂量限值变化均在1％以内。铅门在引入反向、相向运动误差后，其子野面积会相应地增大、减小，靶区剂量变化幅度为≤0.3％、≤0.7%，危及器官剂量变化≤0.5％、≤1.1%。但是在准直器发生同向偏移时，子野面积并不发生变化，靶区和危及器官的变化≤0.3％、≤0.5%。同理，当MLC发生反向、相向位置偏移时，靶区、危及器官的剂量最大改变量可分别达到6％、9％，并且该值随着位置误差的增加而线性增大。当发生同向偏移时，靶区和危及器官的剂量变化均在1％以内。