Proton Arc Therapy Optimization

Abstract: Cancer is ranked among the leading causes of death in the world. During the last decades, the development of advanced cancer treatment software has had an increasingly important role in its treatment. To treat cancer, there are many different therapies, one of which is radiation therapy. In radiation therapy, a central component of the treatment planning software is mathematical optimization of the radiation dose. This thesis concerns proton radiation therapy and aims to propose novel methods for generating advanced treatments using a new technique for delivering the radiation. In conventional proton therapy, the patient is irradiated from a few selected directions, typically two or three. To each irradiating direction corresponds a proton beam whose energy is modulated to control the depth at which the protons deposit their energy. This thesis concerns methods for an alternative technique, called proton arc therapy, where the central idea is to irradiate the patient from significantly more directions, but with fewer energies from each direction. This has the possibility of improving both the outcome and the efficiency of the treatment. The number of energies needs to be constrained due to the fact that they are the major determinant of the treatment delivery time, which is important as it is desired to irradiate the patient for as short time as possible. Thus, the central problem to be solved in proton arc therapy is which energies to use for irradiation at each possible angle. The work in this thesis aims at improving the state of the art by implementing methods for optimally solving this problem. When modeling the treatment as an optimization problem, the objective function is a quantitative evaluation criterion for the delivered dose distribution. In this work, the conventional optimization problem used for proton therapy is extended to proton arc therapy by including a constraint on the number of energies used over the arc. This problem is solved by using different algorithms and heuristics. The methods are evaluated on three different pancreatic tumor cases according to these evaluation criteria: objective function value, treatment delivery time and biological effect of the delivered dose. The developed methods are all able to produce proton arc treatments outperforming conventional treatments with respect to all the evaluation criteria. It is concluded that the proton arc therapy has the potential to outperform conventional proton therapy in all regards. The suggested method to perform the energy selection is a hybrid approach of greedy and reverse greedy algorithms. Future work should focus on the possibility of taking the biological effect into account in the optimization, as well as incorporating machine-specific constraints in the optimization model.