ESTRO 2025 - Abstract Book

S2774

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2025

optimised with RAD and the results used to train a second model (RP-RAD). The validation cases were then optimised with both VMAT and RAD using the RP-RAD model (Fig 1). The RAD technique included a single full arc with one static angle port at gantry 0° and collimator 90°, with collimator angles dynamically optimised throughout the gantry rotation. Plans were tweaked for target coverage.

Plan quality was compared considering: the mean dose to parotid glands, submandibular glands, oral cavity, constrictor muscles, larynx, trachea, and oesophagus. The normal tissue complication probability (NTCP) was estimated using published models [2,3,4]. Wilcoxon signed-rank test was used to compare the plans (Fig 1) in terms of dose and NTCP (p=0.05). Results: Target coverage (D98%≥95%) was fulfilled in all plans. For all OARs the mean dose was lower with RAD compared to VMAT with the lowest dose achieved with RP-RAD optimisation (Fig 2, p<0.05 for all OAR except right parotid). These results were also reflected in the estimated NTCP.