A numerical method to optimise the spatial dose distribution in carbon ion radiotherapy planning

The authors describe a numerical algorithm to optimise the entrance spectra of a composition of pristine carbon ion beams which delivers a pre-assumed dose-depth profile over a given depth range within the spread-out Bragg peak. The physical beam transport model is based on tabularised data generated using the SHIELD-HIT10A Monte-Carlo code. Depth-dose profile optimisation is achieved by minimising the deviation from the pre-assumed profile evaluated on a regular grid of points over a given depth range. This multi-dimensional minimisation problem is solved using the L-BFGS-B algorithm, with parallel processing support. Another multi-dimensional interpolation algorithm is used to calculate at given beam depths the cumulative energy-fluence spectra for primary and secondary ions in the optimised beam composition. Knowledge of such energy-fluence spectra for each ion is required by the mixed-field calculation of Katz's cellular Track Structure Theory (TST) that predicts the resulting depth-survival profile. The optimisation algorithm and the TST mixed-field calculation are essential tools in the development of a one-dimensional kernel of a carbon ion therapy planning system. All codes used in the work are generally accessible within the libamtrack open source platform.