Fast re-planning for radiotherapy by using the reference point method

Abstract

When a patient is diagnosed with cancer, a number of treatments is possible. About half of these patients is treated with radiotherapy. This is a multi-disciplinary field of research, where mathematics, physics and medicine come together. In this thesis the focus will lie on prostate cancer. One of the main problems with this type of cancer is that the organs in the stomach area move, mainly due to volume changes of the bladder. During the treatment, the same treatment plan is used for several weeks, but since these organs move, this plan is not optimal every day. The goal of this thesis is to find a way to make fast optimal plans with the reference point method, such that a new treatment plan can be made every day. The reference point method is used in multi-objective optimisation where the solution obtained is always Pareto optimal. This means that no objective can be improved without worsening at least one other objective. This is due to the fact that the objectives are conflicting. An objective is for example the mean amount of radiation in the rectum. For the reference point method a reference point and an increasing direction of the reference path are needed. Most of the times it is also necessary to add indifference curves. In those cases a sensitivity parameter is needed as well. The goal is to find a good method to choose these values, such that equally good treatment plans can be calculated with the reference point method as with the time-consuming fullmethod, iCycle. Several options for choosing a reference point are discussed. First, the reference point is based on the clinically favourable solutions (iCycle solution) of different scans from the same patient. And then the reference point is based on the iCycle solutions fromscans of different patients. The conclusion was that it is best to use the iCycle solution of the planning-CT of the same patient or the mean of iCycle solutions from other patients as reference point. This is also convenient in practice, since almost no prior data for a specific patient is needed in the treatment planning process and the outliers have little influence.