ESTRO 2020 Abstract Book

S94 ESTRO 2020

4 University Hospital Bern, Department of Radiation Oncology, Bern, Switzerland ; 5 University Hospital Zürich, Department of Radiation Oncology, Zürich, Switzerland Purpose or Objective Daily adaptive proton therapy (DAPT) aims to correct, on a daily basis, anatomical changes and positioning uncertainties. For this, the treatment should be re- optimized on-line as quickly as possible to help minimize the time between daily imaging and treatment delivery, a process that can be achieved in just a few seconds if a simple, but limited accuracy, dose algorithm is used (1). Off-line (i.e. between fractions) however, a slower and more accurate dose calculation, utilizing machine log- files, can be used to accurately reconstruct the delivered dose. This work investigates whether such off-line reconstructed doses, accumulated from previous fractions, can be fed-back into the daily re-optimization process to correct for systematic dose inaccuracies caused by fast on-line optimization, as well as inaccuracies resulting from machine delivery uncertainties. Material and Methods Five fraction treatments for three nasopharynx patients have been simulated, and log-file/Monte Carlo (MC) reconstructed doses of simulated previous deliveries were fed back into a fast on-line plan optimization process, performed using a GPU based ray-casting algorithm (1). Treatment delivery was simulated by adding realistic uncertainties to the planned spot positions and dose was reconstructed using MC (Topas-Geant4). The daily optimization target dose of the N th fraction was then optimized as follows: nominal plan dose and P 0 is the prescribed dose distribution in the target. As such, the optimizer minimizes the difference of the daily plan to this target dose. For evaluation, target homogeneity (D 2% - D 98% / D Prescripton ) and organ at risk D 2% values for accumulated treatment doses have been considered. Results Compared to the nominal plan, degradation of target dose homogeneity in the accumulated dose due to delivery/algorithmic uncertainties was between 1.2 and 3.4% (Fig 1). By feeding back the off-line reconstructed dose into the optimization, homogeneity could be improved by >2% for all three patients. In addition, an average reduction of D 2% for all OARs could be achieved with feedback compared to delivery without adaption (Fig 2). where D i is the delivered dose of fraction i, N 0 is the

Conclusion Dose inaccuracies introduced through the use of a fast and simple dose calculation for daily re-optimization of plans in DAPT and/or due to delivery uncertainties, can be compensated for by including previously reconstructed doses, calculated off-line using log-files and MC, into the daily (on-line) re-optimization process. Although improvements are small (~2%), the approach can mitigate more than half of delivery/dose calculation degradations. The proposed approach is thus a promising candidate for improving treatment accuracy, whilst preserving fast on- line re-optimization. The method still needs to be tested in areas of deforming anatomy however, where dose PD-0191 Knowledge-based planning improves plan quality for high-risk prostate cancer with four dose levels K. Fjellanger 1 , H.E.S. Pettersen 1 , J.A. Hundvin 1 , B. Nygaard 1 , K. Revheim 1 , T.H. Sulen 1 , L.B. Hysing 1 1 Haukeland University Hospital, Department of Oncology and Medical Physics, Bergen, Norway Purpose or Objective Manual VMAT treatment planning for high-risk prostate cancer with four dose levels (SIB) is complex and time consuming, with a number of auxiliary structures and many objectives to be balanced in the optimization. This study investigates if the plan quality can be improved using a system for knowledge-based planning (KBP). Material and Methods 69 patients previously treated with VMAT for high-risk prostate cancer was included in a RapidPlan model created in Eclipse V15.5 (Varian). Prescribed dose was 50, 60, 67.5 and 72.5 Gy in 25 fractions to the pelvic lymph nodes, prostate and seminal vesicles, prostate gland and prostate tumor(s), respectively. Three auxiliary structures were included in the model, in order to define dose limits for parts of the target volumes not overlapping with higher dose level targets. OARs were bladder, bowel and rectum. The dose-volume objectives and priorities were tuned to obtain optimal balance between target coverage and normal tissue sparing in the final model. For 20 patients not included in the model, a RapidPlan (RP) plan was created without any user intervention during the optimization phase, and compared with the manually optimized clinical plan (CP). gEUD was calculated using k accumulation may be more challenging. (1) Matter et al, Acta. Oncol. 2019