Reduced order constrained optimization : application to IMRT planning

Abstract

The major drawback of the method continues to be the time-consuming process of generating treatment plans, which increases when dealing with complex treatment sites such as the lung or head-and-neck. At RPI, a novel method called Reduced-Order Constrained Optimization (ROCO) was previously proposed and applied to a relatively simple cancer site, the prostate, and the results were encouraging. ROCO was able to generate clinically acceptable prostate IMRT plans rapidly.

We also discuss future work, including the continued development of the hot-spot suppressing technique, as well as ROCO as a stepping stone to tackle the complex optimization problem posed by the Volumetric Modulated Arc Therapy (VMAT) radiation technique.

Finally, we introduce a hot-spot suppressing technique that aims to eliminate the current use of non-specific normal tissue structures, generated by the planner to prevent hot-spots outside the PTV. We investigate the incidence of high doses where beams intersect in a lung patient that required several non-specific tissue structures in the clinical plan to control hot-spots. We demonstrate that we can replace all of these structures with a single one that can be defined a priori, and introduced at the constrained optimization stage of ROCO, to suppress all hot-spots, while maintaining satisfactory PTV coverage and OAR sparing.

Next, we develop a prioritized optimization scheme to improve ROCO by eliminating manual repetition of the constrained optimization stage to automatically produce an IMRT plan with target coverage equivalent to, and OAR doses lower than, those in the original plan. The method consists of three stages. First, a baseline ROCO plan with initial constraints for all the OARs is generated. Next, certain critical OAR dose constraints are selected to be further reduced, and the constrained optimization phase of ROCO is re-applied; such constraints are reduced simultaneously until PTV (Planning Target Volume) coverage is violated. Finally, each OAR constraint on a priority list is sequentially pushed until PTV coverage is inadequate. The technique automatically produces superior plans compared to the original ROCO method.

Next, we extend ROCO to the head-and-neck site. Challenges in IMRT planning for this complex site include several levels of prescription and a large, variable number (6-20) of Organs at Risk (OARs). ROCO for head-and-neck generates clinically acceptable plans rapidly and semi-automatically. We provide guidelines for applying ROCO to larynx, oropharynx, and nasopharynx cases, and report the results of a live experiment that demonstrates how an expert planner can save several hours of trial-and-error using the proposed approach.

First, we use ROCO to generate acceptable IMRT plans quickly and semi-automatically for advanced lung cancer patients. Challenges that arise when applying ROCO to this site include the lack of a class solution, a larger treatment site, an increased number of parameters and beamlets, and a variable number of beams and beam arrangement. Our new ROCO implementation has been designed to operate with the treatment planning system and full dose calculation used at Memorial Sloan Kettering Cancer Center (MSKCC).

Intensity Modulated Radiotherapy (IMRT) has revolutionized 3-D treatment planning and conformal therapy by increasing sparing of normal tissue while delivering high radiation doses to tumors. It does this by optimizing the delivery of radiation to irregularly shaped volumes and has the ability to produce concave dose distributions.

This thesis generalizes the ROCO method to two significantly more complex sites, the lung and head-and-neck, as well as improves both its efficiency and efficacy in generating treatment plans that are clinically competitive.