Reduced-Order Constrained Optimization for Rapid IMRT and VMAT Treatment Planning

描述(申请人提供)：调强放射治疗(IMRT)在过去十年中彻底改变了癌症的治疗，因为它可以紧密地调整和增加肿瘤的辐射剂量，同时保护附近对辐射敏感的正常组织，导致比以前的技术更好的局部控制和更少的治疗后并发症。然而，为困难的部位获得临床可接受的调强放射治疗方案的过程仍然极其缓慢，需要忙碌的专家在参数调整的手动反复试验中花费数小时的时间。该项目的目标是使用一种新的自动化方法大幅减少获得临床可接受的调强放疗计划的时间，该方法以计算容易处理和临床有意义的方式直接应用约束优化。假设使用这项技术的临床治疗计划时间将从几个小时减少到几分钟左右。这种新的方法，称为ROCO(降阶约束优化)，将优化和机器学习理论中的成熟概念转化为IMRT计划的新应用，利用了无约束优化的速度和简便性，并引入了降维步骤，使真正的约束优化变得容易处理。该建议的具体目标是(1)将降阶约束优化应用于非小细胞肺癌和鼻咽癌的调强放疗计划，这两种肿瘤的计划过程非常耗时；(2)开发并扩展降阶约束优化范式，将其应用于目前几乎难以计划的前列腺部位调强放疗变种--容积调制弧形疗法(VMAT)；以及(3)将新工具集成到纪念斯隆-凯特琳癌症中心的临床调强放疗计划流程中，使用动力研究验证拟议方法显著提高了计划速度的假设。这些实验将在咨询专业临床治疗规划师和生物统计学家的情况下设计，并使用每个地点约50名患者的匿名数据进行仔细验证。建议的方法的主要好处是大幅减少计划时间，这对于调强放疗和VMAT在临床应用中充分发挥其潜力是至关重要的。在繁忙的诊所，长时间的计划时间会对可用资源造成严重压力，并可能导致治疗延误、接受次优计划，或者在最糟糕的情况下--由于时间压力而出错。从长远来看，所提出的方法将使人们更深入地了解剂量优化问题的关键要素，显著减少当前调强放疗计划的试错性工作特点，并减少治疗方案选择的主观性。 公共卫生相关性：调强放射治疗(IMRT)是一种非常有前景的癌症治疗方法，但目前需要专家治疗规划者数小时的时间才能获得满足医生指定的所有临床限制的计划。这项建议描述了一种新的、自动的治疗计划优化方法，该方法非常快，只需几分钟就可以产生一个直接满足医生所有限制的计划，潜在地为繁忙的诊所节省了大量时间。

DESCRIPTION (provided by applicant): Intensity-modulated radiotherapy (IMRT) has revolutionized the treatment of cancers in the last decade, since it can tightly conform and escalate radiation dose to a tumor while simultaneously protecting nearby radiation- sensitive normal tissues, resulting in better local control and fewer post-treatment complications than previous techniques. However, the process of obtaining a clinically acceptable IMRT plan for a difficult site is still extremely slow, requiring many hours of a busy expert's time in a manual trial-and-error loop of parameter adjustment. The goal of this project is to drastically reduce the amount of time to obtain a clinically acceptable IMRT plan using a new automated method that directly applies constrained optimization in a computationally tractable and clinically meaningful way. The hypothesis is that clinical treatment planning times using this technique will be reduced from several hours to a matter of minutes. The new approach, called ROCO (Reduced-Order Constrained Optimization) translates well-established concepts from optimization and machine learning theory to the novel application of IMRT planning, exploiting the speed and ease of unconstrained optimizations and introducing a dimensionality reduction step that makes true constrained optimization tractable. The Specific Aims of the proposal are to (1) apply Reduced-Order Con- strained Optimization to IMRT planning for non-small cell lung cancers and nasopharynx cancers, where the planning process is highly time-consuming; (2) develop and extend the Reduced-Order Constrained Optimization paradigm to a promising IMRT variant called Volumetric Modulated Arc Therapy (VMAT) for the prostate site, which is currently nearly clinically intractable to plan; and (3) integrate the new tools into the clinical IMRT planning process at Memorial Sloan-Kettering Cancer Center, using a powered study to verify the hypothesis that the proposed method significantly improves planning speed. The experiments will be designed in consultation with an expert clinical treatment planner and biostatistician, and carefully validated using anonymized data from approximately 50 patients for each site. The main benefit of the proposed approach is to drastically reduce planning times, which is critical if IMRT and VMAT are to reach their full potential in clinical application. In a busy clinic, long planning times place a severe stress on available resources, and can result in treatment delays, acceptance of sub-optimal plans or - in the worst case - errors due to time pressure. In the longer term, the proposed approach will provide deeper insight into the critical elements of the dose optimization problem, significantly reduce the trial-and-error effort characteristic of current IMRT planning, and reduce subjectivity in treatment plan selection. PUBLIC HEALTH RELEVANCE: Intensity-modulated radiation therapy (IMRT) is an extremely promising cancer treatment, but currently requires many hours of an expert treatment planner's time to obtain a plan that meets all the clinical constraints specified by the physician. This proposal describes a new, automatic method for treatment plan optimization that is very fast, requiring only minutes to produce a plan that directly meets all the physician's constraints, potentially saving much time for a busy clinic.