ESTRO 2020 Abstract Book

S96 ESTRO 2020

beams shielding the left (Fig. b) and right lung, and (5) various beams to homogenize the dose. Beam weight optimization was performed to achieve the objectives that depend on the fractionation, e.g.: Lung: mean 10 Gy; kidney/brains: mean 12 Gy; rest of body: mean 12 Gy. In vivo dosimetry was performed using MOSFET dosimeters (Best medical, Ottawa, Can.) at the level of the left lung (A and P) and the abdomen (P only).

consortium. The present work aimed to validate this model across the consortium to demonstrate rapid adoption of an available model and to quantify the time savings of using this KBP approach. Material and Methods Five centres (C1-C5) participated in the study, including the two centres who had developed the model (C1 and C2). Each centre generated plans for ten cervical cancer patients using a) standard template-based optimisation (STD), b) a single optimisation using objectives generated by the RP model (RP_1) and c) a plan optimised based on the RP-generated objectives, but with subsequent modifications/interaction as required to produce a clinically acceptable plan (RP_FP). The time taken in each case was recorded. C1 used historical plans for template- based optimisation and hence no timing data was available for these cases. Plans were compared using a consensus set of plan acceptance criteria which incorporated the different metrics used within the centres. Results All centres saw a reduction in planning times when using RP compared to template-based optimisation (see Fig. 1). The model gave a good starting point in all cases, although subsequent optimisations were required to produce a clinically acceptable plan. Despite the need for further optimisations, the time required to generate a final plan was still significantly reduced compared to template-based optimisation for C3-5 (p<0.05), with an average time saving of 75 min. There was no significant difference in the monitor units for RP_FP compared to the STD plan for most centres, with the exception of C3 (RP_FP: 409 MU, STD: 484 MU; p=0.03).

Results From October 2 nd , 2018 to October 2 nd , 2019, 12 patients were treated with our new technique totalling 34 treatment fractions. Treatment planning dose objectives could well be achieved although creating a homogeneous dose distribution in protruding extremities (right upper arm, upper legs) was a challenge and treatment planning took a considerable amount of time (up to 16h). Before treatment, positioning of the patients using the linac light field and MV imager (MV image in Fig. d) could be performed within clinically relevant limits. Results of in vivo MOSFET dosimetry was within the set specifications (+/- 10%): mean deviations: Lung: 6%; Abdomen: -1%. Conclusion We present a TBI technique using an Elekta linac and the Pinnacle TPS that can reliably be used in conditioning of patients prior to stem cell transplantation. This new technique significantly increased the lead time of our TBI treatment with respect to our previous technique that did not use a whole-body CT. However, dose homogeneity improved and dose uncertainty decreased. PD-0193 Validation of a multi-centre knowledge-based planning model for radiotherapy of cervical cancer E. Adams 1 , S. Currie 2 , C. Thomas 3 , A. Pediaditaki 4 , S. Temple 5 , C. South 1 , G. Currie 2 , A. Nisbet 6 1 Royal Surrey NHS Foundation Trust, Radiotherapy Physics, Guildford, United Kingdom ; 2 Beatson West of Scotland Cancer Centre, Radiotherapy Physics, Glasgow, United Kingdom ; 3 Guy's and St Thomas NHS Foundation Trust, Radiotherapy Physics, London, United Kingdom ; 4 Northampton General Hospital NHS Trust, Radiotherapy Physics, Northampton, United Kingdom ; 5 The Clatterbridge Cancer Centre NHS Foundation Trust, Radiotherapy Physics, Liverpool, United Kingdom ; 6 University College London, Department of Medical Physics & Biomedical Engineering, London, United Kingdom Purpose or Objective There are significant resources required for the initial setting up of Knowledge-Based Planning (KBP) models for different treatment sites, particularly as the models depend on a sufficiently sized library of clinically acceptable plans. This can be a barrier for centres with a small number of patients and in rarer clinical sites. A multi-centre group using Varian RapidPlan (RP) KBP, previously reported on their development of a model based on the experience of two of the centres in the

For the PTV, there were no significant differences in D95%. RP_1 gave reduced D99% and increased hotspots for most centres (see Table 1); these were corrected in the final plan. For organs-at-risk, performance of the model was mixed, with variation in both the magnitude and direction of differences in each of the centres. This is likely to be due to differences in local plan acceptance criteria and/or beam energy. Viewed across all centres, there was little difference between the final plan and the template-based optimisation; those differences which reached statistical significance were small and of limited clinical significance.