Degree Name

Doctor of Philosophy


School of Physics, Faculty of Engineering and Information Sciences


As technology advances, so does the quality of treatment offered to cancer patients. Proton therapy, and more specifically proton pencil beam scanning, is currently at the forefront of radiation therapy. Pencil beam scanning offers excellent tumor dose control as well as surrounding organs at risk sparing. Current treatment planning, however, is performed on a static image acquired before treatment. Naturally, this is not a proper representation of the actual patient on a daily basis. Thus, there is a need for adaptive radiation therapy, such as readjusting a given treatment plan based on the patient’s daily setup or a moving tumor location. In order to perform adaptive treatment delivery, appropriate imaging as well as an extremely fast, yet accurate, dose computation engine is needed.

GEANT4 Monte Carlo simulations were performed in order to assess the imaging capabilities and limitations of a proton radiography detector, comparing them to conventional X-ray imaging. In parallel, a small form factor proton radiography system was designed based on available technologies. Thus, photonic bandgap fibers, a CMOS active pixel sensor, and Bicron scintillating fibers were evaluated for proton imaging purposes.

The requisites and limitations of treatment planning for proton pencil beam scanning were further defined, from the acquisition of the treatment planning software’s beam model to the methodologies and treatment robustness. Based on this work, a simplified Monte Carlo algorithm was designed and implemented on the CPU architecture. This computation engine, GMC, was validated against physical observables and then compared to the treatment planning software dose calculation, as well as a ”full” Monte Carlo recomputation.

Proton radiography showed poor spatial resolution but excellent density resolution when compared to X-ray radiography. This density resolution can be of importance when attempting to perform tumor tracking. The lower imaging dose associated with proton radiography is also of interest, especially in pediatric patients. Moreover, the use of a unique beam’s eye view could slightly improve the accuracy of treatment delivery. Photonic bangap fibers, as well as the specific CMOS active pixel sensor used in this work, did not yield promising results for proton imaging. Conversely, Bicron scintillating fibers proved to be suitable for the design of a proton radiography system, as both the individual particle’s position and energy could be acquired.

The treatment planning software’s beam model is very simple, as compared to other modalities. However, the planning stage presented a few limitations, such as a lack of robustness analysis and issues related to spot placement. It was shown that both of these issues could be addressed with the use of a fast, yet accurate, dose computation engine. GMC was successfully implemented on the CPU architecture, and compared extremely well against actual pre-treatment QA measurements. The comparisons against the current algorithm of the treatment planning software and the full Monte Carlo engine matched the expectations for such an algorithm.

The complementing work on proton imaging and fast dose computation algorithm lays a solid foundation to materializing pencil beam scanning adaptive radiotherapy. Future work will focus on generating the necessary synergy between the two systems in order to implement the tools in the clinical setting.