Abstract : Although often quite time consuming, Monte Carlo method is the calculation algorithm that most closely models the actual physics of the energy deposition process. The idea is to use Monte Carlo calculations in daily cancer treatment using radiation to compete with treatment planning systems (TPS) in the delivering of absorbed dose to tumour for specific treatments. To achieve this goal, two points have been particularly studied in this thesis: the validation of the GATE platform for dosimetry applications using electrons, with a specific study concerning ocular brachytherapy treatments using 106Ru/106Rh ophthalmic applicators and the deployment of GATE simulations in a grid environment to reduce the very high computation time of those simulations.
Monoenergetic and polyenergetic electron dose point kernels have been simulated using the GATE platform and compared with other Monte Carlo codes. Three versions of the GEANT4 physics package have been used for the comparisons (5.2, 6.2 and 7.0). Results show that the Multiple Scattering implementation is responsible for the discrepancies observed between the codes. Simulations of ocular brachytherapy treatments compared with Monte Carlo and measurements show a good agreement. The transcription of Hounsfield units from CT images of patient's anatomy to tissue parameters is the other work presented for a next usage of GATE on voxelized images for personnalized dosimetry. The DataGrid and then the EGEE infrastructures were used to deploy GATE simulations to reduce their computation time in order to use them in clinical practice.
The method used to parallelize the GATE simulations is the splitting of the random number generator (RNG) into independent sequences. Computing time tests performed on the grid testbeds show that a significant speed-up is obtained. Functionalities to split, launch and monitor GATE simulations on a grid infrastructure have been implemented on the GENIUS web portal. A first prototype of this portal is accessible from hospital to use the accurate Monte Carlo algorithms in a transparent and secure way for ocular cancer treatments.