Uncertainties in high precision radiation therapy: Impact of deformable image registration methods and of delineation variations on the accumulating dose distribution

近年来，放射治疗技术的发展使剂量应用技术更加精确；因此，现代技术也被称为高精度放射治疗。物理剂量可以在确定的靶体积内精确地应用于肿瘤，同时使邻近器官免受照射的危险。为了达到这种改进的精度，需要进行许多开发。改进的剂量计算和优化方法允许计算机辅助规划复杂的照射计划，也可以覆盖凸形和不规则形状的目标体积。成像设备与照射设备的融合可以减少定位的不确定性，并使放射治疗期间的解剖变化可视化成为可能。今天，图像处理方法有助于在治疗过程中的每个点重新定位靶体积和危险器官；这使得辐照计划能够精确地适应时间变化。然而，使用这些方法，放射治疗链也变得更加复杂。尽管治疗过程中各个步骤的不确定性很小，但在治疗过程中方法的迭代使用过程中，不确定性会累积起来。为了评估整个治疗过程的质量，或者能够干预治疗过程，例如通过采用各种纠正策略，需要考虑累积的不确定性。因此，模拟整个治疗过程并包括发生的不确定性是很重要的。进行这样的模拟仍然是一个挑战。特别是通过可变形图像配准方法来确定时间变量变形是难以验证的。对于相同的输入图像，不同的方法会导致不同的结果；此外，方法的质量取决于使用它们的环境。这些不确定性必须反映在累积剂量分布的不确定性中。本研究项目的目标是将常见的图像配准方法与治疗计划系统相结合。我们计划通过开发一个新的虚拟病人和采用广泛使用的评估方法来研究图像配准方法在剂量分布方面的不确定性。这将使我们能够确定最适合用于放射治疗的图像配准方法，并评估累积剂量分布中不确定性的传播。通过这种方法，我们可以比较竞争性适应性校正策略，或者在计划对先前治疗的患者进行再照射的情况下确定潜在的剩余剂量耐受性。

In recent years, technical developments in radiation therapy have resulted in ever more precise dose application techniques; hence, modern techniques are also referred to as high-precision radiotherapy. The physical dose can be applied precisely to the tumor in the defined target volumes while sparing the adjacent organs at risk from irradiation.Many developments were necessary to achieve this improved precision. Improved dose calculation and optimization methods allowed for the computer-assisted planning of complex irradiation plans that can also cover convex and irregular shaped target volumes. The fusion of imaging devices with irradiation devices allowed for a reduction in positioning uncertainties and made it possible to visualize anatomical changes during radiation therapy. Today, image processing methods help to relocate target volumes and organs at risk at every point during the treatment course; this allows adapting the irradiation plan precisely to temporal changes. However, using these methods, the radiation therapy chain has also become more complex. Even though the respective steps in the treatment course carry only small uncertainties, the uncertainties add up during the iterative use of the methods in the therapy course.To assess the quality of the entire treatment course, or to be able to intervene in the treatment course e.g. by employing various correction strategies, it is required to take the accumulated uncertainties into consideration. Therefore, it is important to simulate the entire treatment course and to include the occurring uncertainties.It is still a challenge to perform such a simulation. Especially the determination of temporal variable deformations through deformable image registration methods is hard to validate. Different methods result in different results for the same input images; in addition, the quality of the methods depends on the context in which they are used. These uncertainties will have to be reflected in the uncertainties of the accumulated dose distribution.The goal of this research project is to integrate common image registration methods with a treatment planning system. We plan to investigate image registration methods regarding their uncertainties in dose distribution by developing a new virtual patient and employing widely used evaluation methods. This will enable us to determine the image registration methods most suited for usage in radiation therapy and to evaluate the propagation of uncertainties in the accumulated dose distribution. With this we can e.g. compare competing adaptive correction strategies, or determine the potentially remaining dose tolerance in case of planning a re-irradiation of previously treated patients.