ESTRO 2022 - Abstract Book

S1462

Abstract book

ESTRO 2022

PO-1663 Evaluation of the dosimetric impact of autodelineation uncertainties in prostate radiotherapy.

K. Alty 1 , D. Marshall 1 , A. Bird 1 , R. Powis 1 , G. Webster 1

1 Worcestershire Oncology Centre, Radiotherapy, Worcester, United Kingdom

Purpose or Objective Several commercial products automatically delineate radiotherapy planning structures, which could improve workflow efficiency. Evaluating these algorithms for clinical use is challenging, with the dosimetric impact of autodelineation uncertainties arguably of most interest. An efficient dosimetric evaluation pipeline has been developed and tested in prostate planning. Materials and Methods 11 Prostate patients had the target volumes manually outlined by the treating clinician 1 . OAR structures (MS) were delineated by Mirada DLCExpert AI autocontouring system (DLC) 2 and then modified manually (DS) by a dosimetrist. An autoplanning script was then run on both structure sets, producing a Mirada-only plan (MP) and a plan based on the modified structures (DP). The dose distributions for both MP and the gold standard DP plans were compared on the DS structures (MP-DS, DP-DS). Time savings due to Mirada were quantified against a historic benchmark. Results All MP-DS plans were clinically acceptable on all metrics and following expert plan review. OAR dose differences are shown in table 1. Bladder dose variations were small, femoral head variations were larger but well within tolerances. Both the rectum and bowel showed larger variations, which disappeared with small modifications to the sup extent of the structures. The average time to modify the autodelineated contours was 6.5 minutes against a dosimetrist-only benchmark average of 14 minutes.

Conclusion Plans optimised using unedited autodelineated OAR structures were found to be clinically acceptable when reported on gold standard outlines. Reporting errors associated with autodelineated structures were small in most cases but warrant further investigation in outlying cases. Production of DS showed over 50% times savings against manual contour creation, with the results of this study suggestive of further gains to be made through a reduction in the modifications needed. 1 RayStation product version: RayStation 9B SP1, with IronPython 2.7. 2 Mirada product version: Workflow Box 2.6 with DLCExpert AI autocontouring, Mirada Medical Ltd., Oxford, United Kingdom

PO-1664 Faster treatment using Halcyon linear accelerator: the experience of Parma University Hospital

M. Maddalo 1 , A. Mazzilli 1 , G. Benecchi 1 , N. D'Abbiero 2 , C. Ghetti 1

1 Azienda Ospedaliero Universitaria di Parma, Medical Physics Department, Parma, Italy; 2 Azienda Ospedaliero Universitaria di Parma, Radiotherapy Department, Parma, Italy Purpose or Objective The Varian Halcyon accelerator could potentially speed up radiotherapy sessions without a significant loss of the overall treatment quality respect to conventional linear accelerators. Patient positioning and treatment setup benefit from the availability of external light-markers and of a simplified and guided workflow. The time dedicated to IGRT can be reduced due by fast CBCT acquisition. In addition, higher maximum speeds of both gantry and MLC could allow to reduce the VMAT beam-on time. This study aims to compare VMAT treatment time between Halcyon and other linear accelerators and to quantify the time saving. Materials and Methods The overall treatment time on the three linear accelerators available at our hospital – Halcyon (H), TrueBeam (TB), 2100 DHX (DHX) - were retrospectively, systematically and anonymously acquired during the first three months of Halcyon activity using InSightive Analytics (IA) software for the following treatment sites: head and neck (H&N), breast, prostate and rectum. IA records the delay between opening and closing of each radiotherapy session at the treatment console. Treatment times