Results of a Pilot Study on Online Adaptive Radiotherapy of Bladder Cancer with Artificial Intelligence-driven Full Re-optimization on the Anatomy of the Day

To evaluate the feasibility and potential clinical impact of online adaptive radiotherapy (oART) of bladder cancer using a novel commercial solution with artificial intelligence (AI) for full re-optimization on the anatomy of the day. A pilot study on oART of muscle invasive bladder cancer was carried out after the clinical implementation of a novel commercial CBCT-based and AI-driven system as the first clinic in the world. Ten bladder cancer patients, with a prescribed dose of 64 Gy in 32 fractions, received one or more fractions as oART, adding up to a total of 100 adaptive sessions delivered over a five-month period. Patient-specific margins were applied after establishing intra-fractional anatomical variations from analyses of CBCT images acquired before and after the first three fractions, and adjusted if needed after any fraction during the course of treatment. Automated treatment planning with full re-optimization to the anatomy of the day, enabled by the AI-driven structure-guided deformable image registration, allowed for margin reductions compared to current standard for image guided radiotherapy (IGRT) of bladder cancer while ensuring target coverage and minimizing dose to organs at risk (OAR). The feasibility of oART for bladder cancer patients was studied, and the observed reductions in treatment volumes and the impacts it had on critical dose parameters to OAR were evaluated. The novel commercial system for AI-driven oART was considered feasible for oART of bladder cancer with AI-generated structures requiring none or minor editing in a majority of the delivered sessions, with a median (Mdn) adaptive process duration of 12.1 minutes (inter quartile range, IQR = 3.0). Adapting to the anatomy of the day enabled a median reduction of the planning target volume of 47 % (IQR = 17 %) compared to current IGRT standard, resulting in a reduction of common dose parameters for the bowel cavity of up to 47 % (V45 Gy: Mdn = 33 %, IQR = 9 %; V30 Gy: Mdn = 21 %, IQR = 8 %). Due to the large inter-fractional target variations observed in bladder cancer patients, the adapted plan was selected for delivery in 97 % of the sessions with the selection criteria being ensured target coverage, reduced dose to OAR, or both. A pilot study on oART of bladder cancer has demonstrated the feasibility of a novel AI-driven solution with a fast-adaptive process enabling reduced treatment volumes. The reduction in dose to OAR indicates a clinical impact by potentially reduced gastro-intestinal toxicity. The pilot study is therefore the basis upon which a clinical phase II study on oART of bladder cancer is currently being designed.