CT Image Estimation for Calculation of Delivered Dose

描述(由申请人提供)： 在前列腺癌的放射治疗期间，通常的做法是在许多天，理想情况下是所有天的治疗中定位前列腺癌，以实现向前列腺癌输送高剂量的目标，同时极大地节省附近对辐射敏感的正常组织。前列腺可以在治疗前立即获得的CT图像中定位，或者通过跟踪标记物在治疗期间植入前列腺进行定位。CT成像的优势在于，图像数据提供了计算自适应放射治疗(ART)所需的实际传输到前列腺和周围组织的辐射剂量的基础，ART是一种定期调整治疗计划以按照最初计划提供最终剂量分布的程序。CT成像的缺点包括金钱、空间和时间开销。标记物跟踪的主要优点包括易用性和在每个剂量部分(例如，10赫兹)期间的频繁采样，潜在地允许动态调整治疗参数。不幸的是，跟踪系统图像数据的缺乏阻碍了ART的实践，对于一些患者来说，可能会排除在前列腺周围使用较小的边缘，以便更好地保留附近的正常组织。这项研究的假设是，在治疗过程中定位的前列腺标志物可以作为将参考CT图像数据映射到治疗空间的基础，以估计治疗前的CT图像数据可以接受，以计算提供的剂量。具体地说，标记将用于在类贝叶斯框架中计算图像匹配项，以优化前列腺的统计可训练可变形形状模型与治疗空间中的标记位置的非刚性配准，该模型包括紧邻周围的组织。根据正在治疗的患者的参考计划图像创建的称为m-rep的患者特定模型将在m-rep独有的与模型相关的坐标系中嵌入底层图像数据，包括与前列腺相关的标记坐标。由配准过程创建以匹配在处理期间定位的标记的变形的m-rep意味着将整个参考图像数据映射到处理空间以估计处理前CT图像的变换。前列腺内部和周围的组织区域被绘制成不同形态的地图。总体目标是建立概念证明，以评估治疗前CT图像可用于计算上述提供剂量的可接受程度：拟议研究的假设是，在放射治疗治疗提供期间定位的前列腺癌标志物可用作将参考CT图像数据映射到治疗空间的基础，以估计可用于计算提供剂量的治疗前CT图像数据。具体地说，标记将用于在类贝叶斯框架中计算图像匹配项，以优化前列腺的统计可训练可变形形状模型与治疗空间中的标记位置的非刚性配准，该模型包括紧邻周围的组织。根据正在治疗的患者的参考计划图像创建的称为m-rep的患者特定模型将在m-rep独有的与模型相关的坐标系中嵌入底层图像数据，包括与前列腺相关的标记坐标。由配准过程创建以匹配在处理期间定位的标记的变形的m-rep意味着将整个参考图像数据映射到处理空间以估计处理前CT图像的变换。前列腺内部和周围的组织区域被绘制成不同形态的地图。总体目标是建立概念证据，以评估治疗前的CT图像可用于计算如上所述的输送剂量。

DESCRIPTION (provided by applicant): During radiation therapy for prostate cancer it is common practice to localize the prostate on many, ideally all, days of treatment to achieve the goal of delivering a high dose to the prostate while greatly sparing nearby radiosensitive normal tissues. The prostate can be localized in CT images acquired immediately prior to treatment or by tracking markers, implanted in the prostate, during treatment delivery. The advantage of CT imaging is that the image data provides the basis for computing radiation doses actually delivered to the prostate and surrounding tissues as needed for Adaptive Radiation Therapy (ART), a procedure for periodically adjusting the treatment plan in order to deliver a final dose distribution as originally planned. Disadvantages of CT imaging include monetary, space and time expenses. Major advantages of marker tracking include ease of use and frequent sampling during each dose fraction, e.g., 10 Hz, potentially allowing dynamic adjustment of treatment parameters. Unfortunately the absence of image data from tracking systems prevents the practice of ART, and for some patients may preclude the use of smaller margins around the prostate that would allow better sparing of nearby normal tissues. The hypothesis of the proposed research is that the prostate markers localized during treatment delivery can be used as the basis for mapping reference CT image data into the treatment space to estimate pre-treatment CT image data acceptable for calculating delivered dose. In particular the markers will be used for computing the image-match term in a Bayesian-like framework to optimize the non-rigid registration of a statistically trainable deformable shape model of a prostate, including immediately surrounding tissues, with marker positions in the treatment space. A patient-specific model, called an m-rep, created from the reference planning image for the patient being treated will embed the underlying image data, including prostate-related marker coordinates, in the model-related coordinate system unique to m- reps. The deformed m-rep created by the registration process to match the markers located during treatment implies a transformation that maps the entire reference image data to the treatment space to estimate the pre-treatment CT image. The tissue region in and around the prostate is mapped diffeomorphically. The overall aim is to establish proof of concept for estimating pre-treatment CT images acceptable for calculating delivered dose as described above PUBLIC HEALTH RELEVANCE: The hypothesis of the proposed research is that prostate markers localized during radiation therapy treatment delivery can be used as the basis for mapping reference CT image data into the treatment space to estimate pre-treatment CT image data acceptable for calculating delivered dose. In particular the markers will be used for computing the image-match term in a Bayesian-like framework to optimize the non-rigid registration of a statistically trainable deformable shape model of a prostate, including immediately surrounding tissues, with marker positions in the treatment space. A patient- specific model, called an m-rep, created from the reference planning image for the patient being treated will embed the underlying image data, including prostate-related marker coordinates, in the model-related coordinate system unique to m-reps. The deformed m-rep created by the registration process to match the markers located during treatment implies a transformation that maps the entire reference image data to the treatment space to estimate the pre-treatment CT image. The tissue region in and around the prostate is mapped diffeomorphically. The overall aim is to establish proof of concept for estimating pre-treatment CT images acceptable for calculating delivered dose as described above.