High-Throughput Computing for a Multi-Plan Framework in Radiotherapy

放射治疗多计划框架的高吞吐量计算

• 批准号: 8077861
• 负责人: WARREN D D'SOUZA
• 金额: $ 29.75万
• 依托单位: UNIVERSITY OF MARYLAND BALTIMORE
• 依托单位国家: 美国
• 财政年份: 2009
• 资助国家: 美国
• 起止时间: 2009-07-01 至 2013-05-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8077861
• 关键词: Accounting; Algorithms; Behavior; Clinical Engineering; Collection; Complex; Complication; Computational Science; Data; Decision Support Systems; Dependence; Development; Dose; Engineering; Environment; External Beam Radiation Therapy; Generations; Genetic Programming; Goals; Health; Intensity-Modulated Radiotherapy; Knowledge; Lead; Linear Accelerator Radiotherapy Systems; Linear Programming; Machine Learning; Maps; Medical; Methods; Modeling; Monte Carlo Method; NIH Program Announcements; Organ; Patients; Physicians; Physics; Process; Property; Radiation; Radiation Oncology; Radiation therapy; Relative (related person); Research Personnel; Risk; Sampling; Shapes; Simulate; Solutions; Speed; Structure; Surface; System; Technology; Time; Toxic effect; base; cluster computing; combinatorial; computer science; computerized; computing resources; direct application; graphical user interface; heuristics; improved; innovation; insight; novel; novel strategies; predictive modeling; process optimization; programs; research clinical testing; tool; treatment planning; tumor

DESCRIPTION (provided by applicant): Computerized planning for radiation delivery via either external beam radiation therapy (EBRT) or intensity- modulated radiation therapy (IMRT) from linear accelerators is a complex process involving a large amount of input data and vast numbers of decision variables. Such large-scale combinatorial optimization problems are typically intractable for conventional approaches such as the direct application of the best available commercial algorithms, and thus specialized methods that take advantage of problem structure are required. Radiation treatment planning (RTP) problems are further complicated by the fact that they are multi-objective, that is, the RTP optimization process must take into account a trade-off between the competing goals of delivering appropriate doses to the tumor and avoiding the delivery of harmful radiation to nearby healthy organs. The goal of this proposal is to harness distributive computing via the Condor system for High Throughput Computing (HTC) within an RTP environment. The specific aims for this proposal are: 1) To develop a Nested Partitions (NP) framework that guides a global search process for optimal IMRT delivery parameters using HTC. 2) To develop parallel HTC-based linear programming (LP) methods to efficiently solve the dose optimization problem in IMRT for each given set of beam angles or beam apertures. (3) To exploit a high-throughput computing (HTC) environment and the developed NP/LP/segmentation framework to efficiently generate multiple plans for each given patient case. (4) To couple this multi-plan framework with a decision support system (DSS) that includes planning surface models, a graphical-user-interface (GUI) and machine learning tools to prediction OAR complication in order to aid in the ranking and selection of the generated treatment plans. This proposal requires a multi-disciplinary approach that is best conducted within the framework of the Innovations in Biomedical Computational Science and Technology program announcement. It brings together an interdisciplinary team of investigators with expertise in medical physics, mathematical programming, industrial engineering and clinical radiation oncology that is crucial to the development of the proposed multi- plan framework using HTC in radiation therapy. PUBLIC HEALTH RELEVANCE: The goal of this proposal is to develop a multi-dimensional platform for sophisticated treatment planning of radiation delivery. It will develop novel algorithms that will enable generation of superior treatment plans with the added advantage of increasing the speed of treatment planning. Further, it will allow physicians to know beforehand the quality of the treatment plan relative to the multiple treatment objectives and be able to determine the treatment complication scenario beforehand.

描述（由申请人提供）： 用于经由来自线性加速器的外部射束放射治疗（EBRT）或调强放射治疗（IMRT）的放射递送的计算机化规划是涉及大量输入数据和大量决策变量的复杂过程。这种大规模的组合优化问题通常是棘手的传统方法，如直接应用最好的商业算法，因此需要专门的方法，利用问题的结构。放射治疗计划（RTP）问题由于它们是多目标的事实而进一步复杂化，也就是说，RTP优化过程必须考虑向肿瘤递送适当剂量和避免向附近健康器官递送有害辐射的竞争目标之间的权衡。该建议的目标是利用分布式计算通过秃鹰系统的高吞吐量计算（HTC）在RTP环境中。该提案的具体目标是：1）开发一个嵌套分区（NP）框架，该框架引导使用HTC的最佳IMRT输送参数的全局搜索过程。2)针对给定的射束角度或射束孔径，开发基于并行HTC的线性规划（LP）方法，以有效地解决调强放射治疗中的剂量优化问题。(3)利用高通量计算（HTC）环境和开发的NP/LP/分割框架，为每个给定的患者病例有效地生成多个计划。(4)将该多计划框架与决策支持系统（DSS）耦合，该决策支持系统（DSS）包括计划表面模型、图形用户界面（GUI）和机器学习工具，以预测OAR并发症，从而帮助对生成的治疗计划进行排序和选择。该提案需要采用多学科方法，最好在生物医学计算科学与技术创新计划公告的框架内进行。它汇集了一个跨学科的研究人员团队，他们在医学物理，数学编程，工业工程和临床放射肿瘤学方面具有专业知识，这对在放射治疗中使用HTC的拟议多计划框架的发展至关重要。公共卫生关系：该提案的目标是开发一个多维平台，用于复杂的放射治疗计划。它将开发新的算法，使上级治疗计划的生成具有增加治疗计划速度的额外优势。此外，它将允许医生预先知道治疗计划相对于多个治疗目标的质量，并且能够预先确定治疗并发症情况。