Evaluating the quality of radiotherapy treatment plans with uncertainty using data envelopment analysis

Stubington, Emma and Ehrgott, Matthias and Nohadani, Omid (2020) Evaluating the quality of radiotherapy treatment plans with uncertainty using data envelopment analysis. PhD thesis, Lancaster University.

[thumbnail of 2020EmmaStubingtonPhD]
Text (2020EmmaStubingtonPhD)
2020EmmaStubingtonPhD.pdf - Published Version
Available under License Creative Commons Attribution-NonCommercial-NoDerivs.

Download (2MB)


External beam radiation therapy is a common treatment method for cancer. Radiotherapy is planned with the aim of achieving conflicting goals: while a sufficiently high dose of radiation is necessary for tumour control, a low dose of radiation is desirable to avoid complications in normal, healthy, tissue. This thesis aims to support the radiotherapy treatment planning process for prostate cancer by evaluating the quality of proposed treatment plans relative to previous plans. We develop a variable selection technique, autoPCA, to select the most relevant variables for use in our Data Envelopment Analysis (DEA) models. This allows us to evaluate how well plans perform in terms of achieving the conflicting goals of radiotherapy. We develop the uncertain DEA problem (uDEA) for the case of box uncertainty and show that for small problems this can be solved exactly. This study of uncertainty is motivated by the inherently uncertain nature of the treatment process. Robust DEA, uDEA and simulation are applied to prostate cancer treatment plans to investigate this uncertainty. We identify plans that have the potential to be improved, which clinicians then replan for us. Small improvements were seen and we discuss the potential difference this could make to planning cases that are more complex. To aid this, we develop a prototype software, EvaluatePlan, that assesses the efficiency of a plan compared to past treatment plans.

Item Type:
Thesis (PhD)
ID Code:
Deposited By:
Deposited On:
24 Feb 2020 16:55
Last Modified:
30 Jun 2024 23:52