ESTRO 2022 - Abstract Book

S643

Abstract book

ESTRO 2022

Conclusion The AI proposed in this study was able to automatically select beam angles for proton liver treatment planning that were close to the human-selected ground truth. Treatment plans using these AI-selected angles achieved comparable dose distributions to those using human-selected angles. In conjunction with knowledge-based optimization methods this could be a tool to efficiently automate proton treatment planning in liver lesions.

PD-0731 Integrated beam angle optimization in IMRT autoplanning for lung cancer

K. Fjellanger 1,2 , L.B. Hysing 1,2 , B.J.M. Heijmen 3 , H.E.S. Pettersen 1 , I.M. Sandvik 1 , T.H. Sulen 1 , S. Breedveld 3 , L. Rossi 3

1 Haukeland University Hospital, Department of Oncology and Medical Physics, Bergen, Norway; 2 University of Bergen, Institute of Physics and Technology, Bergen, Norway; 3 Erasmus University Medical Center, Department of Radiotherapy, Rotterdam, The Netherlands Purpose or Objective IMRT with fixed beams is a much applied technique for treatment of locally advanced non-small cell lung cancer (LA- NSCLC), to avoid large lung volumes receiving low dose. However, manually selecting the optimal beam configuration for each patient is a challenging task due to large anatomical variations between patients. The aims of this study were to use automated IMRT planning with integrated beam angle optimization (BAO) to 1) enhance plan quality compared to manual planning, 2) investigate patterns in selected beam configurations and 3) investigate the impact of the applied number of beams on plan quality. Materials and Methods 26 LA-NSCLC patients were prospectively included. The prescribed dose was 60-70 Gy in 2 Gy fractions. All patients had a clinical 6-beam IMRT plan (CLIN), manually created by an expert planner. Using a novel in-house developed system for automated multi-criterial IMRT planning with integrated BAO, clinically deliverable plans with 4, 6 and 8 optimized beams were created for each patient (AUTO). Candidate beam angles were 140°-40° for right-sided tumors and 320°-220° for left- sided tumors (5° spacing). Dose-volume parameters related to toxicity and patterns of selected beam angles were compared for 6-beam CLIN and AUTO plans. 4-, 6- and 8-beam AUTO plans were mutually compared. The Wilcoxon signed-rank test ( p ≤ 0.05) was used for statistical testing. Results In the CLIN plans, there were mainly small variations from the planning beam template and all patients had opposing beams in the AP direction. In contrast, the angles in the 6-beam AUTO plans were spread out across the candidate beam space, demonstrating more patient-specific selection (Fig. 1). The resulting AUTO plans were dosimetrically clearly superior to the CLIN plans (Fig. 2). While PTV coverage and lung dose were similar, the median heart D mean was reduced from 9.0 Gy to 8.1 Gy (p = 0.02), median esophagus D mean from 20.3 Gy to 18.5 Gy (p = 0.02), median heart V 30Gy from 11.0% to 6.2% (p = 0.002) and median esophagus V 20Gy from 38.4% to 36.8% (p = 0.008). Fig. 1b shows large differences in selected beams between 6-beam CLIN and AUTO plans for an example patient, with clear impact on heart and esophagus sparing. Dosimetric QA at the linac proved deliverability of the 6-beam AUTO plans. Overall, increasing the number of optimized beams improved OAR sparing, with a larger impact of going from 4 to 6 than from 6 to 8 (Fig. 2).