Motion Management of Pancreatic Cancer in MRI-Guided Radiotherapy

描述(由申请人提供)： 意义：内脏运动是胰腺癌放射治疗的最大技术挑战之一。人们已经对分割内肿瘤运动进行了广泛的研究。这项研究提供了量化方法，并使新的治疗方法能够减轻运动引起的不良剂量学效应。与广泛研究的肺部肿瘤运动相比，在肿瘤周围有放射敏感的系列器官的情况下，提高胰腺治疗的准确性具有同等或更重要的临床意义，保留这些器官将显著提高我们向肿瘤输送更有效剂量的能力。然而，由于胰腺软组织X线和CT对比度差的技术挑战，胰腺癌治疗中的器官运动管理严重不发达。基准标记物不常放置，当放置时，不能充分描述复杂的多器官运动。创新：MRI引导的放射治疗有可能克服这些挑战，利用基于器官距离的门控放射治疗，而不是预先选择的呼吸时相。然而，为了使其能够进行这样的胰腺运动管理，需要提高MRI的采集速度，以便能够以足够的质量采集动态体积图像以用于器官勾画。此外，还没有开发出用于临床运动管理的快速消化治疗前和治疗期间MRI图像的方法。个体化运动边缘和门控放射治疗都需要明确的器官分割，但手动勾画大量的成像帧是不切实际的。胰腺的自动分割工具还没有开发出来，但迫切需要。我们将通过利用患者解剖学的空间和时间一致性来获得具有足够质量的运动量化的加速3D MRI。我们将开发一种新的流形聚类约束字典学习(MCDL)方法来有效地分割MRI图像，并为胰腺解剖提供准确的运动评估。我们假设，改进的运动监测将显著提高肿瘤剂量和周围正常器官的保留率。目的：1.开发从欠采样k空间数据中获取三维动态图像的方法。2.通过优化和验证MCDL方法，开发了一种自动分割预期获得的加速MRI图像的过程。3.使用准确描述的胰腺肿瘤运动来量化剂量学增益。在核磁共振引导的放射治疗机(ViewRay)上对建议的门控计划进行测试，并进行稳健性和交付能力测试。影响：局部晚期胰腺癌患者预后不佳，但局部完全缓解的患者中位生存期可显着提高。该项目的成功将定义更准确的患者特定运动范围，并促进门控放射治疗，从而显著减少关键器官剂量，增加肿瘤剂量，以获得更大的完全局部应答率。该项目开发的方法也将适用于其他肿瘤，如宫颈癌和肝癌。

DESCRIPTION (provided by applicant): Significance: Internal organ motion is one of the greatest technical challenges for pancreatic cancer radiation therapy. Extensive research has been conducted on intrafractional tumor motion. This research has provided quantification and enabled novel treatment methods to mitigate motion induced adverse dosimetry effects. Compared to widely studied lung tumor motion, improving pancreas treatment accuracy has equal or greater clinical importance where the tumor is surrounded by radiosensitive serial organs, sparing of which will significantly improve our ability to deliver more effective doses to the tumor. However, organ motion management in the pancreatic cancer treatment is severely underdeveloped due to technical challenges related to poor soft tissue X-ray and CT contrast of pancreas. Fiducial markers are not commonly placed and when placed, inadequately describe complicated multiple-organ motion. Innovation: MRI guided radiotherapy has the potential to overcome these challenges utilizing gated radiotherapy based on organ distances instead of a pre-selected breathing phase. However, to enable it for such pancreatic motion management, MRI acquisition speed needs to be increased so dynamic volumetric images can be acquired with sufficient quality for organ delineation. Furthermore, methods to rapidly digest both pre- and during-treatment MRI images for clinical motion management have not been developed. Both individualized motion margin and gated radiotherapy require explicit organ segmentation but manual delineation of the large number of imaging frames is impractical. Automated segmentation tools for pancreas have not been developed but urgently needed. We will acquire accelerated 3D MRI with sufficient quality for motion quantification by exploiting the spatial and temporal coherence of patient anatomy. We will develop a novel manifold clustering constrained dictionary learning (MCDL) method to efficiently segment the MRI images and provide accurate motion assessment for pancreas anatomy. We hypothesize that the improved motion monitoring will result in significantly improved tumor dose and surrounding normal organ sparing. Aims: 1. Develop methods to acquire 3D dynamic images from under-sampled k-space data. 2. Develop a process to auto-segment prospectively acquired accelerated MRI images by optimizing and validating a MCDL method. 3. Quantify dosimetric gains using accurately described pancreatic tumor motion. Test and robustness and deliverability of the proposed gated plans on a MRI-guided radiotherapy machine (ViewRay). Impact: Patients with locally advanced pancreatic adenocarcinoma have a dismal prognosis but the median survival can be significantly improved for patients who have complete local response. Success of the project will define more accurate patient specific motion margins and facilitate gated radiotherapy that significantly reduce critica organ doses, increase tumor doses for greater complete local response rates. Methods developed by this project will also be applicable to other tumors such as the cervical and liver cancer.