Radiation therapy is a complex process where a given target volume receives a given dose of radiation divided over one or multiple treatments. Every step in this process can introduce some types of uncertainties into the problem which may compromise the quality of the treatment. Typically, a volume larger than the actual tumor is irradiated to make the treatment more robust against these uncertainties. This comes at the cost of normal tissue irradiation and an increased risk of toxicity. In this dissertation, we investigate approaches to managing uncertainties in radiation therapy treatments for lung cancer patients.

In the first part of the dissertation, we focus on the process of designing a treatment plan which involves selecting appropriate beam angles and deciding the right amount of radiation dose to the tumor cells, while sparing the normal tissue surrounding the tumor. Selecting the optimal set of treatment beam angles, called beam angle optimization (BAO), involves a very large-scale combinatorial optimization problem with many local minima. In order to identify an efficient approach to obtain high quality beam angles, we first examine the strengths and weaknesses of some existing BAO optimization methods including both global and local search algorithms. We then propose a hybrid framework to overcome some of the weaknesses observed in these methods.

Next, we perform an in-depth study into the impact of interplay effect, which results from relative motion of the tumor and proton beam, on the dose distribution in the patient with lung cancer. The dynamic dose distribution, that provides an estimation of delivered dose under the influence of interplay effect, is calculated by simulating the machine delivery processes on the moving patient described by 4D computed tomography (4DCT) during the dose delivery process by linking timestamps of each on/off switch of proton spots, spills, energies, and fields to patient respiratory cycles. We introduce a clinically applicable metric for clinicians to use for determining the magnitude of the uncertainties caused by interplay effects. We then explore the techniques of fractionation and iso-layered re-scanning for mitigating these interplay effects.

In the last part of the dissertation, we develop a robust adaptive optimization framework for intensity modulated radiation therapy (IMRT) for lung cancer, where temporal variation of tumor volume and its associated uncertainties throughout the course of the treatment are accounted for to re-optimize the treatment plan for the following sessions. This framework gives an insight into the trade-off between sparing the healthy tissues and ensuring that the tumor receives a sufficient dose. With this trade-off in mind, we demonstrate that our robust adaptive solution outperforms a non-adaptive solution and a nominal (no uncertainty) solution on a clinical case.