Monte Carlo-based Inverse Treatment Plan Optimization for Intensity Modulated Proton Therapy

Performance of a hybrid Monte Carlo‐Pencil Beam dose algorithm for proton therapy inverse planning

“…This is significantly bigger than the factor 5 achieved by Li et al. with their adaptive particle sampling scheme. Note that in our case, this factor was computed by dividing the time employed to compute the accurate dose influence matrix in the full MC approach () by the time needed to compute the accurate forward dose used in the correction matrix ().…”

“…The first inner loop ( t = 0) must achieve an optimal approximate dose in order to have a good starting point for the subsequent corrections ( t > 0). Therefore, a convergence criterion (3) based on the relative dose difference between two consecutive iterations and was always used for this initial run:where represents the total number of voxels summed over all ROIs involved in the objective function and is the dose prescription for the target volume (PTV). Optimization was stopped when had not exceeded 0.005% for the last five iterations.…”

“…The first inner loop (t = 0) must achieve an optimal approximate dose in order to have a good starting point for the subsequent corrections (t > 0). Therefore, a convergence criterion (3) based on the relative dose difference between two consecutive iterations k and k À 1 14,15 was always used for this initial run:…”

“…To cope with this issue, Y.Li et al . have recently developed an adaptive particle sampling (APS) scheme that iteratively performs MC dose computation and plan optimization.…”

“…To cope with this issue, Y.Li et al 14,15 have recently developed an adaptive particle sampling (APS) scheme that iteratively performs MC dose computation and plan optimization. However, this approach needs a substantial change in the optimization workflow and perfect integration of the dose engine during the optimization process.…”