Deformable Image Registration and Reconstruction

最佳的图像引导自适应放射治疗需要患者解剖结构的4D表示，其允许通过生物成像、计划和模拟、近距离放射治疗的递送和每个IMRT部分的管理的过程来跟踪肿瘤和正常组织体素的位置。该项目的科学目标是研究非刚性图像配准的新方法，用于构建和验证患者解剖结构在治疗过程中发生变化时的表示。实际目标是创建一套图像处理资源，使图像引导自适应放射治疗技术在临床上的常规应用。在具体目标1中，我们将研究轮廓驱动的可变形配准方法，用于将骨盆中的高剂量近距离放射治疗（HDR）剂量分布映射到IMRT剂量分布，以及用于将生物图像配准到外部射束计划图像，包括开发一种新型的表面匹配算法，该算法考虑了轮廓不确定性。为了有效地将信息从计划CT图像映射到机载CT图像，在管理每个每日部分之前采集，我们将开发不需要手动轮廓标志的快速参数化图像变形算法。具体目标2。我们将研究通过将变形模型匹配到平面图像投影序列来从不完整投影数据重建CT图像的新方法，从而将图像重建和变形配准集成到单个过程中。这将用于开发具有改进的时间分辨率的患者呼吸的4D解剖表示，并从更高的时间分辨率估计分次内解剖变形。 分辨率序列的2D图像。最后，在具体目标3中，将开发用于估计可变形图像配准的不确定性和误差的新方法。

Optimal image guided adaptive radiotherapy requires a 4D representation of the patients anatomy, that allows the position of tumor and normal tissue voxels to be tracked through the processes of biological imaging, planning and simulation, delivery of brachytherapy, and administration of each IMRT fraction. The scientific objective of this project is to investigate novel methods of nonrigid image registration for constructing and validating such representations of the patient's anatomy as it changes during the treatment process. The practical goal is to create a suite of image processing resources that will enable the routine application of image-guided adaptive radiotherapy techniques in the clinic. In specific aim 1, we will investigate contour-driven deformable registration methods for mapping high-dose brachytherapy (HDR) dose distributions in the pelvis to IMRT dose distributions, and for registering biological images to external beam planning images, including development of a novel surface matching algorithm that accounts for contouring uncertainties. To efficiently map information from planning CT images to onboard CT images, acquired prior to administering each daily fraction, we will develop fast parametric image deformation algorithms that do not require manually contoured landmarks. In Specific Aim 2. we will investigate novel methods for reconstructing CT images from incomplete projection data by matching deformation models to sequences of planar image projections, thereby integrating image reconstruction and deformable registration into a single process. This will be used to develop 4D anatomic representations of patient respiration with improved temporal resolution and to estimate intrafraction anatomic deformation from higher temporal resolution sequences of 2D images. Finally, in Specific Aim 3, novel methods for estimating the uncertainty and error of deformable image registration will be developed.