The Role of Automation in Radiotherapy Treatment Planning for Prostate Cancer

Abstract

Each year in the UK, almost 50, 000 men are diagnosed with prostate cancer and of these approximately 30 % receive external beam radiotherapy (EBRT) as part of their treatment (Cancer Research UK, 2018). Treating prostate cancer patients therefore compromises a significant proportion of a typical EBRT centre’s total clinical workload. It has long been acknowledged that changes to the contents of the bladder, bowel and rectum can cause the prostate to move considerably within the pelvic cavity (Moiseenko et al., 2007; Hosni et al., 2017). To ensure that the malignant disease receives the prescribed EBRT dose under this positional uncertainty, a margin of healthy tissue that includes radiosensitive organs at risk (OAR) surrounding the prostate is generally also treated. Adaptive radiotherapy (ART) can ameliorate this situation by adapting the treatment to account for day-to-day changes in pelvic anatomy thereby allowing the disease to be targeted more accurately and reducing the irradiated volume (Nijkamp et al., 2008; Antico et al., 2019). Unfortunately, although ART techniques provide dosimetric and potential clinical benefits to patients, clinical adoption of ART is far from widespread. ART brings additional work to the radiotherapy treatment planning process, which is magnified over the large prostate cancer patient population such that it quickly becomes unmanageable in most centres. Automation therefore has an important role to play in ART workflows for prostate radiotherapy and hence is the focus of this thesis. Chapter 1 outlines and compartmentalises the current non-ART prostate radiotherapy treatment planning pathway and presents a literature review of published attempts to automate it. Subsequent chapters present individual studies that build on these published works and culminate in a fully-automated knowledge-based treatment planning workflow. Rigorous analyses of treatment plans generated using this fully-automated workflow show that they are generally of at least comparable quality to manually generated clinical treatment plans – although gross inaccuracies in auto-contouring of anatomical structures can be a limitation. Nevertheless, the proposed fully-automated workflow could provide significant efficiencies for treatment planning departments, which could be exploited to aid with the implementation of ART techniques and provide patients with earlier access to their cancer treatments. Publication of the fully-automated knowledge-based treatment planning workflow and the data used to drive it would also allow treatment planning information and expertise developed over more than a decade at The Christie NHS Foundation Trust to be shared easily and quickly with other centres worldwide.

Date of Award	1 Aug 2021
Original language	English
Awarding Institution	The University of Manchester