Using collected patient’s data to configure treatment plans for new patients in radiotherapy

Abstract

Nowadays around 50% of all patients diagnosed with cancer are treated with radiotherapy. Radiotherapy uses ionizing beams to irradiate the tumour. This radiation damages and destroys cells, such that it is impossible for this cell to grow and divide. When the tumour is irradiated, the surrounding healthy tissues gets damaged as well. The goal is to destroy the tumour cells, while saving the healthy tissue as much as possible. Radiotherapy is a multi-objective optimization from a mathematical point of view. The different organs are the objectives and the dose in these organs must be minimized. Decreasing the dose in one organ influences the dose in all other organs. This gives an infinite range of optimal solutions. Clinically, such an optimal solutions does not exist. We are searching for a clinically desirable solution which is a good compromise between all objectives. To find clinically desirable solutions, the Erasmus MC has developed Erasmus-iCycle, an algorithm for automated multi-objective radiotherapy treatment plan optimization. Unfortunately, configuring this algorithm is highly time-consuming, as the essential information only known by the physicians is not explicitly available. However, this information is present in existing treatment plans of previously treated patients. This project focuses on extracting that knowledge to configure Erasmus-iCycle, so it can treat new patients in a similar fashion. The aim of this project was to develop a method which uses knowledge of previous patients and reproduces this input to validate the feasibility of the method. In this thesis a number of methods is discussed, the advantages and disadvantages are compared. The conclusion we found was that the best method to solve this problem is the simplest method. This method brings us very close to the old solutions, it is simple to implement and the computation time is very short.