ESTRO 2023 - Abstract Book

S265

Saturday 13 May

ESTRO 2023

Conclusion As expected, both models perform favourably on data that is reflective of their training population. Each model performed comparably on the external UK dataset. The results suggest that clinical utility can be found in bespoke and externally developed models. However to better understand performance and generalisability, independent testing should be recommended for institutions or vendors developing autosegmentation models for radiotherapy. Model evaluation on the test set alone insufficient to assess performance and generalisability, particularly in a public health setting. PD-0331 Automated dynamic trajectory radiotherapy planning using 4pi-IMRT path-finding - proof of concept H.A. Loebner 1 , J. Bertholet 1 , S. Mueller 1 , G. Guyer 1 , D. Frei 1 , W. Volken 1 , O. Elicin 1 , D.M. Aebersold 1 , M.F.M. Stampanoni 2 , M.K. Fix 1 , P. Manser 1 , P. Mackeprang 1 1 Inselspital, Bern University Hospital and University of Bern, Division of Medical Radiation Physics and Department of Radiation Oncology, Bern, Switzerland; 2 ETH Zuerich and PSI, Institute for Biomedical Engineering, Villigen, Switzerland Purpose or Objective Dynamic trajectory radiotherapy (DTRT) adds dynamic table and collimator rotation during beam-on to state-of-the-art volumetric modulated arc therapy (VMAT). In this work, we propose a novel automated DTRT treatment planning process (auto-DTRT) which reduces planner bias in treatment planning in terms of DTRT path determination and manual intensity optimization by combining 4pi-IMRT based path-finding and automatic intensity modulation optimization (AIO). Materials and Methods The auto-DTRT workflow consists of four steps, automated using scripting on a research version of the Eclipse treatment planning system: First, a 4pi-IMRT plan is created with 100 fields, set up equidistantly on a sphere around the target (Figure 1A). Second, the fluence of the 4pi-IMRT plan is optimized using a lexicographic based AIO, which uses a treatment site-specific wish-list of optimization objectives as input. This wish-list is derived from internal clinical guidelines and groups minimal optimization goals for the target and organs-at-risk (OARs). Third, the AIO-optimized fluence of each beam multiplied by their respective MLC opening area is mapped onto a gantry- table map. On this map, one DTRT path is determined by an A* path-finding algorithm and duplicated with a jaw collimator split (Figure 1B&C). Dynamic collimator rotation is obtained by minimizing the width between opposing MLC leaves. Fourth, another AIO is performed with the same wish-list, this time for the obtained DTRT paths, followed by the final dose calculation (Figure 1D&E). The auto-DTRT process is explored for a clinically motivated unilateral postoperative case of Adenoid-Cystic carcinoma in the left parotid gland. Local irradiation was performed with 2 Gy per fraction to 50 Gy (elective) and 66 Gy (sequential boost). The resulting plan was compared to a manually planned VMAT plan of that case. Plan quality is assessed according to clinical goals by a clinician. Computation time is recorded.