Intensity Modulated Proton Therapy Optimization Under Uncertainty: Field Misalignment and Internal Organ Motion
MetadataShow full item record
Intensity modulated proton therapy (IMPT) is one of the most advanced forms of radiation therapy, which can deliver a highly conformal dose to the tumor while sparing the dose in healthy tissues. Compared to conventional photon-based radiation therapy, IMPT is more flexible in delivering radiation dose according to different tumor shapes. However, this flexibility also makes the optimization problems in IMPT harder to solve, e.g., it requires larger memory to store data and longer computational time. Furthermore, proton beams are very sensitive to different uncertainties, such as setup uncertainty, range uncertainty and internal organ motion. These uncertainties can greatly impact the quality of clinical treatment. Therefore, this dissertation aims to investigate different optimization methods for treatment planning and to handle a variety of uncertainties in IMPT. First, to solve the fluence map optimization (FMO) problem in IMPT, we propose a method to formulate the FMO problem into a molecular dynamics model. So that, the FMO problem can be optimized according classical dynamics system. This method combines the advantages of gradient-based algorithms and heuristic search algorithms. Next, we develop and validate a robust optimization method for IMPT treatment plans with multi-isocenter large fields to overcome the dose inhomogeneity problem caused by the setup misalignment in field junctions. Numerical results show that the robust optimized IMPT plans create a low gradient field radiation dose in the junction regions, which can minimize the impact from misalignment uncertainty. Compare to conventional techniques, the robust optimization method leads the whole treatment much more efficient. Lastly, we focus on a two-stage method to solve the beam angle optimization (BAO) problem in IMPT with internal organ motion uncertainty. In the first stage, a $p$-median algorithm is developed for beam angle clustering. In the second stage, a bi-level search algorithm is used to find the final beam angle set for the treatment. Furthermore, Support vector machine (SVM) is used for beam angle classification to reduce the search space and the 4D-CT information is incorporated to handle the internal organ motion uncertainty. Results show that the two-stage BAO method consistently finds a high-quality solution in a short time.