The protontherapy is an innovative technique for cancer treatment in critical areas, such as the eye or the head. Even though the interaction of protons with human tissues is a well-known physical phenomenon which gives rise to the protontherapy, there are uncertainties on the proton trajectory due to heterogeneities in the irradiated tissue, the calculation of the beam parameters in the planning treatment affects the theoretical benefits of the protons and the chosen dose delivery process. Thus, methods for irradiation quality control have been suggested. Most of them rely on utilizing the mapping of the positron emitters generated during the irradiation. They are detectable and quantifiable thanks to the use of the PET (positron emitter tomography), a medical imaging technique mainly used for the cancer expansion assessment. PET acquisitions were proposed and then realized on phantoms and patients after protontherapy. The quality control relies on comparing the measured radioactive distribution to the simulated β+ distribution. The modeling of the positronic activity generated by protons in the irradiated area can be divided into three steps: the simulation of the proton beam, the modeling of the proton interactions in the irradiated object and the modeling of the PET acquisition. Different ways of simulating these steps are possible. This PhD work suggests different ways of modeling the three steps and evaluates theirs benefits for the irradiation quality control. We have restrained our evaluation to the verification of the proton range and to the uncertainties related to the proton range. This research work utilizes on irradiations in homogenous and inhomogeneous areas in a head model. We have compared the uncertainties on the proton range measured thanks to the following β+ distributions: 1) A β+ distribution obtained by modeling the irradiation with a proton beam simulated analytically and simulated using the complete Monte Carlo method; 2) A Monte Carlo modeling of the proton range using the GEANT4 software (versions 9.2 and 9.4) relying and using cross-sections; 3) A simulation of the PET acquisition using a simplified modeling and a Monte Carlo modeling. Our results show that a simplified modeling of the beam does not affect the estimation of the proton range. Besides, the Monte Carlo modeling of the PET camera enables modeling the noise present in the PET signal measured in a homogeneous area. Preliminary results of the PET camera modeling are presented in a head model (inhomogeneous). Finally, a simplified modeling of the PET camera enables evaluate the proton range in a homogeneous area with a 1mm-precision, which is equivalent to the reproducibility of the PET offline measure as described in (Knopf et al., 2008).
