ESTRO 2025 - Abstract Book

S4321

RTT - Treatment planning, OAR and target definitions

ESTRO 2025

Conclusion: Across treatment, variations in BV had limited impact on overall cumulative dosimetry. Dosimetric criteria were out of tolerance for the target and OARs on some fractions. Further work is required to determine the optimal frequency and timing for adaptation.

Keywords: bladder volume, cervix, dosimetry

727

Proffered Paper Promoting the practical clinical integration of automated radiotherapy planning: validation across multiple institutions and disease types Lei Yu 1,2 , Jiazhou Wang 1,2 , Weigang Hu 1,2 1 Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China. 2 Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China Purpose/Objective: Automated treatment planning (ATP) in radiotherapy based on deep learning (DL) holds great promise by predicting the best achievable dose distribution on a novel patient anatomy, but in clinical practice, the output plan is usually less flexible to achieve individualized trade-offs [1]. To promote its practical integration into clinical workflow, we proposed a hybrid strategy by combining DL dose prediction and individualized clinical goals in a commercial TPS (uTPS, United Imaging Healthcare, Shanghai, China) to generate directly deliverable plans. A retrospective study across multiple institutions and disease sites were conducted to evaluate the performance of the proposed ATP solution. Material/Methods: Based on a channel attention densely-connected U-Net architecture [2], DL models of five disease sites, including head & neck, lung, breast, cervix, and rectum, were trained separately based on historical datasets of a single institution (A) and tested for clinical use among three institutions (A, B, and C). A prioritized clinical goal list, representing trade-offs between target coverage and organ sparing, was established for each site as a reference to