On the Disconnect Between the Modeling and Evaluation in Radiation Therapy

Shojaei, Sayyedeh Nastaran

doi:https://doi.org/10.21985/n2-nxry-n718

Work

On the Disconnect Between the Modeling and Evaluation in Radiation Therapy

Public

Download PDF

Cancer radiation therapy relies on optimized treatment plans whose quality assessment is judged by dosimetric planning aims. It is computationally prohibitive to incorporate the planning aims into the optimization models. Therefore, there exists a disconnect between the two steps of (1) optimizing a plan and (2) evaluating the optimized plan, requiring a trial-and-error decision-making process. This disconnect leads to three main problems, namely that the decision-making process provides no guarantee for optimality, does not permit standardization and comparability (knowledge transfer), and is time-consuming. In fact, data shows that final plans almost always violate some of the recommended planning aims, indicating that clinical considerations often require deviations from planning aims. In this dissertation, we study this disconnect from both modeling and evaluation points of view. In Chapter 1, the radiation therapy planning process is reviewed and four phases in the planning process are described. Next, the disconnect in the radiation therapy planning process is discussed. Finally, we present an overview of this dissertation. In Chapter 2, we focus on the evaluation step. A framework is derived to optimize planning aims by taking into account both the biophysical constraints to safeguard clinical toxicity levels and the planning constraints that we extract from past treatment data. In addition, we enable individualization of plans by constructing robust aims, i.e., when oncologists deviate from aims, the outcomes remain unaffected and optimal. This decision support framework can serve hospitals to establish their guidelines by leveraging technology with the goal of preventing errors (improves outcome), promoting consistency (improves institutional learning and alignment among organizations), and increasing efficiency (improves patient flow and reduces costs). In Chapter 3, we apply this framework to head-and-neck cancer cases and use a large set of past treatments to extract planners' knowledge. The resulting optimized aims have the advantage that they (a) are more consistent, (b) require fewer relaxations, (c) enforce higher-quality solutions, and (d) are robust to potential deviations. While the optimized aims are developed by international agencies as a guideline for the planners, they do not guarantee that, for a given patient, the planner meets the optimized aims. Therefore, in Chapter 4, we develop an additional tool to minimize the deviation from the planning aims by taking into account both recommended aims and the patients' information. In addition, we assist the planners to make clinically acceptable solutions by generating multiple robust plans with small deviations. This tool can be integrated into current treatment planning systems to assist planners in the modeling phase by leveraging data-driven optimization algorithms. Utilizing this tool is expected to lead to a planning process with higher quality and time efficiency.

Creator