ESTRO 2023 - Abstract Book

S243

Saturday 13 May

ESTRO 2023

25 beams have been audited at the time of writing. For the basic tests we observe that 23/25 beams are within 2 % for the beam output with 14/25 beams being within 1 %. For 1 centre the difference between measurements and TPS calculations was higher than 3 %, which is under investigation. The differences are within 3 % for the other basic tests normalised to the measured beam output. Figure 1 and Figure 2 show the alanine and films results for the E2E SBRT test respectively. For the E2E delivery the alanine results for 15/25 beams are within 3 % of the calculated doses, with 4 beams being within 1 %. Only for 1 beam did the measured dose deviated by more than 5 % from the calculations. For the film through the tumour the constraints of the gamma analysis were 5%/1mm. 23/25 beams had a passing rate > 95 % with 20/25 beams having > 98 %. 2 beams had a passing rate between 90 % - 95 %. The film op top of the lungs was analysed with 3%/2mm as criteria. 20/25 beams had a passing rate > 95 % with 12/25 having > 98 %. 3 beams had a passing rate < 90 %.

The comparison of the alanine/EPR dosimetry with IAEA was within 0.5 % for all tested beams.

Figure 1: Alanine/EPR results for the E2E SBRT delivery for 25 beams.

Figure 2: Gamma analysis results for the E2E SBRT delivery for 25 beams.

Conclusion Although dose-to-medium is mainly used in clinic, Acuros Dw had to be used here to respect the calibration of the detectors. The difference is only 2%. The results of the basic dosimetry and the complex dosimetry for the SBRT delivery are very satisfactory. MO-0308 AI-based radiotherapy treatment planning quality assurance: A multi-institutional study P. Kalendralis 1 , S.M. Luk 2 , R. Canters 3 , D. Eyssen 3 , A. Vaniqui 3 , C. Wolfs 3 , L. Murrer 1 , W. van Elmpt 1 , A.M. Kalet 4 , A. Dekker 1 , J. van Soest 5 , R. Fijten 6 , C.M. Zegers 1 , I. Bermejo 3 1 Department of Radiation Oncology (Maastro), GROW School for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands, Department of Radiation Oncology (Maastro), Maastricht, The Netherlands; 2 The University of Vermont Medical Center, Burlington, Vermont, United States, Radiotherapy department, Vermont, USA; 3 Department of Radiation Oncology (Maastro), GROW School for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands, Department of Radiation Oncology (Maastro), Maastricht , The Netherlands; 4 Department of Radiation Oncology, University of Washington Medical Center, Seattle, United States, Department of Radiation Oncology, Washington, USA; 5 Brightlands Institute for Smart digital Society (BISS), Faculty of Science and Engineering, Maastricht University, Heerlen, The Netherlands, Brightlands Institute for Smart digital Society (BISS), Heerlen, The Netherlands; 6 Department of Radiation Oncology (Maastro), GROW School for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands, Department of Radiation Oncology (Maastro), Department of Radiation Oncology (Maastro) , Maastricht , The Netherlands Purpose or Objective Artificial intelligence (AI)-based applications have potential to assist physicists and technicians in routine clinical tasks such as quality assurance (QA) of radiotherapy treatment planning. For instance, the Bayesian network (BN) alert system developed by Luk et al.(1) has shown to be a promising tool for the detection of potential treatment planning errors. The goals of this study were to 1) assess the effectiveness of an evolved version of the BN with new variables and links in an international multi-centric setting, and 2) establish an interoperable framework that works across different technologies, clinical guidelines, and patient characteristics. Materials and Methods Treatment planning, diagnostic and dose prescription data from three different radiotherapy centres were collected: Maastro (Netherlands) using ARIA, University of Washington (UW) and University of Vermont Medical Center (UVMMC) using Mosaiq as “record and verify systems”. Data from the three centres were harmonised and discretised, in order to predict