ESTRO 2021 Abstract Book

S1454

ESTRO 2021

Results In test set, the accuracy of the classification model was 95.5%. The MAE were 0.92, 1.12, 2.26, 2.52 and the RMSE 1.52,1.73,3.21,3.60 for 3%/3mm, 3%/2 mm, 2%/3 mm, 2%/2 mm, respectively. The trend and position of predicted dose difference were consistent with the measured dose difference. Conclusion Virtual QA based on Unet++ can be used to classify the field passed or not, predict gamma pass rate under different gamma criteria, and predict dose difference. The results show that Unet++ based Virtual QA is promising in quality assurance for radiotherapy. PO-1730 Feasibility of Virtual Reality-based target volume delineation for craniospinal irradiation M. Volpini 1 , A. Jooya 1 , D. La Russa 2 , J. Sutherland 2 , R. Samant 1,3 , V. Nair 1,4 1 University of Ottawa, Radiation Oncology, Ottawa, Canada; 2 Ottawa Hospital Research Institute, Radiation Oncology, Ottawa, Canada; 3 University of Ottawa Research Institute, Radiation Oncology, Ottawa, Canada; 4 University of Ottawa Research Insitute, Radiation Oncology, Ottawa, Canada Purpose or Objective Accurate and specific target delineation constitutes the vital element in any successful radiation therapy treatment. Current treatment planning platforms require users contour slice-by-slice, which is time consuming and labour intensive. Automated segmentation methods require imaging gradients or high contrast structures in images, post processing and still the final contour may not be patient specific as created from previous datasets. Virtual Reality (VR)-based target delineation platforms may possess the added advantage of volumetric contouring as well as the ability to contour in planes beyond the three conventional cardinal planes. We report the first clinical use and feasibility of a VR-based target delineation platform for target volume delineation in support of craniospinal irradiation. Materials and Methods Five pediatric patients with resected standard risk medulloblastoma were retrospectively selected for this ethics-approved feasibility study. For each patient, a three-dimensional (3D) planning CT with 3-mm slice thickness was used for contouring. DICOM structure sets were imported into the VR target delineation software, Elucis, (Realize Medical) as well as the Monaco TM treatment planning system (Elekta). CTV_Cranial and CTV_Spinal were contoured as per the SIOPE consensus guidelines for medulloblastomas in both platforms. Semi-automation tools including interpolation tools were available to use as needed. Each contour was generated by the same resident in both platforms and were reviewed by a staff radiation oncologist for validity. The total time to contour each structure was compared. In addition, contour similarity was compared using objective parameters. Results The average duration for VR-based targeting was 16 ± 2 min and 33 ± 3 min for the CTV_Cranial and CTV_Spinal structures, respectively, compared to 26 ± 3 min and 57 ± 4 min in Monaco. This represents a 40% reduction in the time required to contour using VR for both structures, or roughly 35 ± 5 minutes saved. It was noted that ready access to 3D visualizations and oblique 2D cross-sections in the VR platform aided in the target delineation process. Conclusion VR-based contouring significantly reduces the time required to delineate complex brain and spinal target volumes without sacrificing accuracy. Additional work is warranted to fully characterize the potential of VR- based contouring of CSI cases, and for other complex treatment sites. PO-1731 Streamlined Quality Assurance on Positioning Guidance Systems with Single Phantom Setup Y. Du 1 , S. Zhou 1 , J. Li 1 , S. Yu 1 , H. Yue 1 , M. Wang 1 , H. Wu 1,2 1 Peking University Cancer Hospital & Institute, Department of Radiation Oncology, Beijing, China; 2 Peking University Health Science Center, Institute of Medical Technology, Beijing, China Purpose or Objective This study was to propose and validate a streamlined quality assurance (QA) program which can be efficiently performed with a single phantom setup to check performances of positioning guidance systems including robotic couch, x-ray modalities (kV-kV, MV-MV, CBCT), surface guidance system (AlignRT), lasers and optical A pseudo-patient treatment plan based on the AlignRT cube phantom was designed and approved. After the cube was randomly set up on the couch, the initial position offsets were simultaneously acquired in AlignRT and CBCT. After the 6DoF couch shift was applied to restore the cube to its reference position, the residual offsets were acquired by kV-kV pair, MV-MV pair and AlignRT in the same process, where the couch motion accuracy can be derived as well. Before wrapping-up, laser alignment to the cube marker lines were checked and ODI values were recorded. The QA program has been approved as a regular weekly QA test on a VitalBeam linac, and we analyzed the results over 50 weeks for clinical validation. distance indicator (ODI). Materials and Methods