ESTRO 2025 - Abstract Book

S3518

Physics - Optimisation, algorithms and applications for ion beam treatment planning

ESTRO 2025

adaptation hardly feasible. We present here a beamlet-free[1] adaptative workflow that dynamically combines Monte Carlo dose calculation with the spot weights optimization process.

Material/Methods: The proposed adaptive workflow is illustrated in Figure 1. The beamlet-free algorithm iteratively simulates proton batches for spots sampled from a probability distribution, evaluates their impact on the objective function, and updates the distribution using stochastic coordinate descent. The algorithm accounts for the dose already delivered and leverages the initial plan P 0 by using the optimized spots displaced with the deformation field, preserving plan’s original information as much as possible to limit changes. The method was initially tested on multiple water phantom cases. To further validate the workflow and explore its potential, we tested it on two lung cases and a cervix case. For the lung patients, the initial reference plan was optimized on the first 4DCT phase. Delivery was simulated phase by phase, starting with partial delivery on the first phase. For each subsequent phase, the plan was adapted to the new anatomy before continuing delivery, resulting in nine reoptimizations across the 10 phases. For the cervix case, the CT scans of two consecutive fractions with significant anatomical changes were used, with the first CT to create the reference plan and the second for adaptation halfway through the treatment delivery. Reference and adapted plans were not robustly optimized, using the in-house open-source treatment planning system OpenTPS[2].

Results: For all cases, CTV coverage criteria is met with a D98 at 58.61Gy and 58.37Gy for the two lung cases (Dp=60Gy) and 43.97Gy for the cervix case (Dp=45Gy). Figure 2 shows, for one patient, that the dose delivered to OARs is comparable to the initially planned dose, maintaining sufficient OAR sparing.