ESTRO 38 Abstract book

S1105 ESTRO 38

The ITV's volume was raised by an average of 26% [8.1%- 44.6%].The difference in ITV mean dose was 3.91% [0.2%- 6.2%] , D2% was 0.5% [0.2%-4.5%], D50% was 4.2% [1%- 6.7%], D95% was 7.2% [3.3%-11.4%] and D98% was 8.3% [3.7%-12.1%]. Differences in dose distribution were less important for patients with less variation in volume of ITV Conclusion In this study 4D-CBCT based dose calculations was mainly affected by limited CBCT image quality. Improvement of CBCT image quality is necessary to accurate dose calculations. EP-2017 GANs covert CBCT to CT for head-neck, lung and breast: paired vs unpaired; single-site vs generic M. Maspero 1 , M.H.F. Savenije 1 , T.C.F. Van Heijst 1 , A.N.T.J. Kotte 1 , A.C. Houweling 1 , J.J.C. Verhoeff 1 , P.R. Seevinck 1 , C.A.T. Van den Berg 1 1 UMC Utrecht, Radiation Oncology, Utrecht, The Netherlands Purpose or Objective CBCT offers a representation of daily anatomy that may be used for online dose calculation and adaptation. However, dose calculations cannot be performed on CBCT due to lack of HU calibration, limited FOV and presence of image artefacts. This study investigates the use of generative adversarial networks (GANs) to convert CBCT into CT. Such networks allow very fast image conversion and thus facilitate online adaptation. However, CBCT to CT conversion using paired learning is problematic given the possible anatomical interscan differences breaking the “paired” (PA) assumption of data consistency. Here, we investigate the use of unpaired (UP) training, which obviates the need for consistent pairwise datasets (e.g. registered, anatomically matched) in the training. In this explorative study, we investigated the performance of PA vs UP learning and compared site-specific vs generic trained network. Material and Methods CBCTs of 88 patients diagnosed with head-neck (HN, 31), lung (29) and breast (28) cancer undergoing radiotherapy were rigidly registered according to the clinical procedure and resampled to the planning CT (XVI, Elekta). PA vs UP To perform PA training, we used a conditional GAN (cGAN), while for UP training, we used a cycle-consistent GAN (cycleGAN). The two networks were trained in 2D transverse planes mapping CBCT to CT Hounsfield Units. For each anatomical site, the networks were trained on 15 patients (training set) and evaluated on the remaining patients (test set). Single vs generic To verify whether a single network could be used for all the patients independently of the anatomical site, we trained both the cGAN and the cycleGAN on the data of 45 patients. Image comparison in terms of mean absolute error (MAE) and mean error (ME) in the FOV of the CBCT vs planning CT was performed on the test set for both the experiments. Results cGAN and cycleGAN training required about 1 and 5 days, respectively, on a GPU Tesla P100 (NVIDIA). Forward evaluation took about 20 s. PA vs UP For all the three sites, discontinuities between 2D transverse slices were more visible after PA compared to UP training (Fig1). In the case of UP training, some residual image artefacts were present in the transverse plane, especially for the breast cases. Mean MAE and ME were for UP and PA are reported in Fig2. Single vs generic For all the anatomical sites, training with all the patients resulted in mean MAE and ME within 1σ respect to training on patients of each site (Fig2).

Figure 1 . Distribution of correction vectors applied for each patient and over all patients (blue boxplot). According to clinical decision Patient 9 (P9) were positioned not applying rotations. No correlation between the 3D translation correction vector and patient weight and age was found (r<0.3). No statistically significant difference between 3D correction vector in lower (P1-P7) and upper (P8-P12) limb was found (pvalue=0.07) while it was significantly patient-dependent (pvalue<<0.001). Conclusion The use of multiple combinations of immobilization devices for limb-extremities allowed for relatively small setup errors. This report might contribute to an Institutional standardization of limb-extremities positioning guidelines. EP-2016 Evaluation of 4D cone beam CT-based dose calculation for SBRT lung cancer treatment S. Bellefqih 1 , B. Benadon 1 , A. Roque 1 , N. Gaillot 1 , S. Servagi-Vernat 1 1 Institut Jean Godinot, Radiotherapy, Reims, France Purpose or Objective Stereotactic body radiation therapy (SBRT) has become a standard treatment for patients with medically inoperable early-stage non-small cell lung cancer (NSCLC). Daily image-guidance is crucial to ensure correct patient set-up prior to treatment delivery, with 3D-CBCT the conventional method used in image guidance. More recently, 4D-CBCT has been used to take into account tumor motion induced by respiration. Another potential use of 4D-CBCT is to determine the actual dose received by the tumor during treatment by using 4D-CBCT intra- fraction. Material and Methods This study included five patients treated with stereotactic body radiotherapy. Radiation treatments were delivered on an Elekta Versa HD linear accelerator. Prior to each radiotherapy treatment fraction, a 4D-CBCT was performed using XVI 4.5. Symmetry (from Elekta package of software solutions for IGRT). All translational errors were corrected prior to the treatment and then a 3D-CBCT was acquired to measure the residual error. Finally, a per- treatment 4D-CBCT was acquired to verify intra-fraction motion. The intra fraction 4D-CBCT was used to delineate the GTV, and the ten GTV merged to form an ITV. A 5-mm margin was added to define a planning target volume (PTV). Finally, the initial plans were recomputed on the CBCT images using a patient specific stepwise curve (Hounsfield units to density). Internal target volumes (ITVs; D98%, D95%, D2%) were compared between simulation CT–derived treatment plans and 4D-CBCT- based plans. Results