Denoising Monte Carlo doses for fast and accurate proton therapy treatment planning

Proton therapy is one of the main radiation therapy treatments nowadays. Using the spread out Bragg peak property, we can irradiate a tumor while saving healthy tissue around it. In order to design a radiation therapy plan, we need to simulate the dose distribution on a CT patient image. The algorithm that performs this task is based on the Monte Carlo method. However, Monte Carlo simulation suffers from statistical noise due to its stochastic nature. This noise can be reduced by simulating more protons. Unfortunately, there is a trade-off between noise reduction and duration of the simulation. Decreasing the noise significantly increases the computing time. In this thesis, we propose a solution to reduce the calculation time and maintain a low amount of statistical noise. In other words, we propose to simulate fewer protons and introduce a correction term to the optimization algorithm.