ESTRO 2020 Abstract book

S997 ESTRO 2020

with use over time. Finally, the application of this solution to all solid cancers is feasible as this challenge is fundamentally a detection and localization task based on routine imaging. Future work is underway in the anatomies of brain, head and neck, breast, kidney, and lung cancer. PO-1711 A CBCT based quality assurance technique for an MR-only radiotherapy workflow for head patients S. Irmak 1 , L. Fetty 1 , T. Nyholm 2 , G. Heilemann 1 , D. Georg 1 , N. Nesvacil 1 , P. Kuess 1 , W. Lechner 1 1 Medical University of Vienna / Vienna General Hospital, Department of Radiation Oncology, Vienna, Austria ; 2 Umeå University, Department of Radiation Sciences, Umeå, Sweden Purpose or Objective Deep learning techniques showed remarkable performance for image-to-image translations, such as for the generation of synthetic CTs (sCT) from MRs for MR-only treatment planning. However, the performance of a neural network is directly linked with the number of images used for training and the variation of training and test datasets. The performance can be described during the training process with the test dataset, but is difficult to define for patients without a planning CT (pCT) in an MR-only workflow. This study proposes the use of CBCT acquisition to evaluate the sCT performance in terms of dose distribution accuracy for patients without a planning CT. Material and Methods A pretrained CycleGAN network was used for the MR-to- sCT conversion. Dose differences calculated on the sCTs and the corresponding CBCTs were investigated for 11 patients suffering from brain lesions. VMAT plans were created using the RayStation treatment planning system. The automatic density overriding method (CBCT RS ) implemented in RayStation and a population-based dose calculation method (CBCT pop ) were employed for CBCT based dose calculation. Dose distributions were analyzed and compared using the Gamma analysis with a 2%/2mm criteria. A Kruskal Wallis test was applied to assess parameter differences for significance, assuming significance level of 5%. Further, differences in DVH metrics such as D mean for the PTV and D 2% for the organs at risk like brain, brain stem, optic chiasm, optic nerves and lenses were compared. Results Using the pCT as reference, the mean gamma-pass-rates were 99.0±0.4% for sCTs. Mean gamma pass rate values were 99.0±0.8% and 99.1±0.8% for the CBCT RS and CBCT pop , respectively. The mean gamma-pass-rates comparing sCT and CBCT resulted in 99.2±0.6% and 98.4±1.6% for CBCT pop and CBCT RS , respectively. The differences between the gamma-pass-rates of the sCT and two CBCT based methods were not significant. D mean for the PTV deviated about 0.3±0.5% for sCT, 0.9±0.5% for CBCT RS and 0.2±0.7% for CBCT pop compared to pCT based dose calculation. D 2% for the critical organs deviated less then 1.1% compared to pCT for all image sets. Comparison of the sCT based dose distributions and CBCT based dose distributions lead to the D mean deviations for PTV of about 1.2% for CBCT RS and 0.1% for CBCT pop , while the D 2% metrics for organs at risk deviated less than 1.4% for CBCT RS and less than 0.7% for CBCT pop . Conclusion Results demonstrate good agreement between sCT and CBCT based calculations with the pCT. Given a proper CBCT calibration is applied, CBCT based dose calculations can be used for quality assurance procedures in an MR only radiotherapy workflow for head patients, for example to highlight geometrical variations in the sCT conversion.

PO-1712 Dosimetric Impact of MRI inaccuracy in SBRT in the presence of motion: digital phantom study T. Torfeh 1 , R. Hammoud 1 , S. Aouadi 1 , S. Paloor 1 , N. Al- Hammadi 1 1 National Center for Cancer Care & Research, Radiation Oncology, Doha, Qatar Purpose or Objective Magnetic Resonance Imaging (MRI) is increasingly being used in radiotherapy applications for tumor delineation and tracking in the presence of respiratory motion. The purpose of this work is to investigate the impact of magnetic resonance imaging geometric distortions on dose distributions for liver cancers during respiratory motion for an MRI-only radiation therapy planning. Material and Methods An in-house motion platform and a control point based phantom were used to calculate the MR distortion map during motion for a Cine FIESTA sequence which is the standard sequence for target delineation in liver cancer RT. The calculated distortion map was used to deform original sets of images issued from a digital phantom generated from two liver cancer patients as shown in figure 1. Fig. 1 a) 3D and 2D views of the digital phantoms extracted from a) the first patient and b) the second patient. Liver, heart, spinal cord, kidneys, lungs and the tumor are shown. Targets of different sizes were created and automatically delineated on the original and distorted datasets. Using Eclipse TM treatment planning system, treatment plans were optimized on the distorted dataset. These plans were then transferred to the original dataset and the dose distribution was analyzed. A highly conformal volumetric modulated arc therapy (VMAT) technique was utilized to irradiate the target. Dose Volume Histograms (DVH) including D 50 , D min and D mean were benchmarked for accuracy measurements. Results The mean magnitude of the geometric distortion was 0.7, 0.9 and 1 mm for radial distances of 50, 100 and 150 mm respectively. Blurring was observed during motion causing an increase in the Full Width at Half Maximum FWHM of the objects of ≈30%. Our results showed that the dose received by the original set of images is higher than the dose received by the distorted images on which the plan was initially optimized. Mean differences in the DVH parameters for the distorted and undistorted treatment plans ranging between -1 and - 14 % and -1.8 and -15% was observed for the target and the OARs respectively. Results also showed that the difference in the DVH is sensitive to the size and position of the target as shown in figure 2. Higher differences were recorded as the distance to the isocentre increases and/or the tumor size decreases.