S1985
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
Results: Figure 2 illustrates the results. Compared to
radiation therapy. Conventional robust optimization mitigates these effects but does not explicitly control the probability of meeting the planning goals. We propose a probabilistic optimization framework that explicitly models systematic and random uncertainties over full treatment courses and directly targets a specified goal fulfillment probability. Material/Methods: Each treatment course is modeled as a sequence of fractions affected by a systematic error and random per-fraction errors. The treatment course dose is the sum of the fraction doses. In the probabilistic optimization, we simulate multiple treatment course scenarios and aim to fulfil the goals for a fraction of these. To this end, we generalize the probabilistic formulation of Bohoslavsky et al. [1] into a quantile- trimmed power mea n function which enables scaling between conditional expected value and quantile optimization. For computational efficiency, doses are computed for preselected points, and interpolation is used to approximate doses for arbitrary shifts [2], see Figure 1.The method was evaluated on a prostate case treated with VMAT and subject to systematic and random setup uncertainties. The probabilistic optimization considered 1000 simulated treatment course scenarios and aimed for a 90% probability of a minimum CTV dose of 77 Gy using two uncertainty modes: systematic/random σ =0.28/0.0 cm and σ =0.2/0.3 cm, both corresponding to a 0.7 cm margin according to the Van Herk rule [3].The plans were compared to conventional planning using a 0.7 cm margin. The robustness evaluation used 1000 independently sampled treatment course scenarios considering systematic/random σ =0.2/0.3 cm.
conventional planning, probabilistic optimization achieved higher probability of clinical goal fulfillment for the OARs and reduced average dose to the external, bladder, and rectum, while maintaining target coverage. The CTV minimum dose goal (D99 ≥ 76.23 Gy) was slightly less strict than the optimized goal (D100 ≥ 77 Gy) and was fulfilled in 100% of the scenarios. On average, the OAR goal fulfillment increased from 59% for the conventional method to 69% and 80% for the probabilistic methods with σ =0.28/0.0 cm and σ =0.2/0.3 cm, respectively.Considering random errors in the optimization led to less homogeneous dose distributions, with the CTV D2 goal fulfilled in 91% of the scenarios compared to 100% for the other methods.
Conclusion: Probabilistic optimization enables improved control over clinical goal fulfillment and can improve OAR sparing at maintained target coverage probability. References: [1] R. Bohoslavsky, M. G. Witte, T. M. Janssen, and M. Van Herk. Probabilistic objective functions for margin- less IMRT planning. Physics in Medicine & Biology, 58(11):3563, 2013.[2] A. Fredriksson, E. Engwall, and B. Andersson. Robust radiation therapy optimization using simulated treatment courses for handling deformable organ motion. Physics in Medicine & Biology, 66(4):045010, 2021.[3] M. Van Herk, P. Remeijer, C. Rasch, and J. V. Lebesque. The probability of correct target dosage: dose-population histograms for deriving treatment margins in radiotherapy.
Made with FlippingBook - Share PDF online