Integration of motion perturbations in a comprehensive robustness evaluator based on a fast Monte Carlo dose engine for PBS proton therapy

Souris, Kevin [UCL] Barragan Montero, Ana Maria [UCL] Lee, John Aldo [UCL] Sterpin, Edmond [UCL]

Background: For ensuring safe treatments with proton therapy delivered by pencil beam scanning (PBS), it is essential to evaluate the robustness of the treatment plans against numerous uncertainties, among which the conversion of Hounsfield Units into stopping powers, setup errors, and breathing motion. For this purpose, we have developed a tool enabling comprehensive robustness evaluation, including against variable breathing motion patterns. Methods: In order to assess the robustness of the PBS plan, multiple scenarios of treatment realizations are simulated by randomly sampling uncertainties from probability distributions reported in the literature. Uncertainty scenarios are then modeled by manipulating the planning 4DCT series to simulate realistic daily treatment images: 1.Setup errors are simulated by translating CT images 2.Stopping power errors are simulated by scaling mass densities in the images 3.Variation of the motion amplitude is simulated by generating new 4DCT phases. To do so, velocity fields are first calculated by registering 4DCT images to a reference phase. The concept of velocity field used by Janssens et al (2010) facilitates field manipulation and ensures the physical consistency of deformation. Deformation amplitude is modified by scaling the velocity fields. A new 4DCT series is then created by deforming back the reference image to each phase, employing the scaled fields. Dose distributions are calculated on each simulated 4DCT using the fast Monte Carlo code MCsquare. Its 4D dose calculation algorithm enables the simulation of the interplay between tumor and treatment beam motions. Variable motion periods are simulated in the robustness analysis. Results: To validate the amplitude variation model, a 40% increase of the initial breathing amplitude was simulated for a lung tumor case. The motion is then analyzed in the generated 4DCT and compared to the initial 4DCT. An effective variation of 39% of the tumor displacement was measured. The treatment plan was obtained by using a 4D robust optimizer developed in-house (MIROpt), without considering breathing amplitude variations. Its robustness was verified using the presented method. The robustness test results, reported as DVH-bands, revealed a sufficient plan robustness against treatment uncertainties typically reported in the literature. However, without repainting, the interplay effect significantly degraded the dose distributions. Conclusions: The developed model enables realistic robustness evaluation of proton therapy treatments, including for mobile targets. Most uncertainty models apply to both photon and particle treatments. Moreover, the generated 4DCTs can be directly used as uncertainty scenarios to feed robust optimizers in commercial TPS.