ESTRO 38 Abstract book

S1141 ESTRO 38

Western General Hospital-, Department of Oncology Physics, Edinburgh, United Kingdom Purpose or Objective The accuracy of radiotherapy dose calculation from cone beam computerised tomography (CBCT) depends on estimation of the electron density accounting for beam hardening, the effects of metal artefacts and inhomogeneous object scatter. However, this is made challenging by the inherent sampling deficit in CBCT and the desire to minimise patient dose. Several studies have considered this problem from the point of view of reconstruction, registration or a combination of the two. Here we present the enhanced electron density and dose planning accuracy made possible by a new direct quantitative reconstruction method called Polyquant. Material and Methods The stereotactic end-to-end verification (STEEV) phantom patient, Model 038, was used for this study (CIRS, Norfolk, VA, USA). Using a 1 mm head protocol planning fan beam CT images were acquired on a Philips Brilliance Big Bore CT scanner fitted with a flat couch (Philips, Amsterdam, The Netherlands). Treatment plans were prepared using the Eclipse treatment planning system (Varian Medical Systems, Inc., Palo Alto, CA, USA) for a 200 MU 10x10 cm 2 field delivered at 9 gantry angles and the dose calculated at a predetermined reference point. The STEEV phantom was set up on a Varian Truebeam linear accelerator and CBCT images acquired using the on-board imager with the preset head acquisition protocol. Physical dose measurements were acquired using a Pinpoint ionization chamber (PTW, Freiburg, Germany) inserted into the dosimetry cavity of the STEEV phantom. Using the ‘raw’ CBCT projections standard Feldkamp, Davis and Kress (FDK) reconstruction was carried out. Quantitative reconstruction was also performed using our new Polyquant method. Dose calculation was carried out on the FDK-CBCT and the Polyquant-CBCT images on Eclipse and compared to the physical measurement. Results Quantitative results from the physical measurements on the STEEV phantom compared to the calculated dose by Eclipse using the planning CT, FDK-CBCT and Polyquant- CBCT are shown in Table 1 and Figure 1. Not only is the error reduced from the FDK-CBCT result, but the Polyquant reconstruction significantly outperforms the planning CT accuracy with a 26.5% reduction in error. This may be a surprising result due to the usual superiority of fan-beam CT, from its heavily reduced scatter over CBCT, and the routine use of this modality in practice for planning. We would expect that if we had access to the raw data from the fan-beam CT also, the Polyquant applied to this would likely result in better accuracy still.

Table 1: Results from calculated and physical measurements. Conclusion These preliminary results show that a significant increase in CBCT dose calculation accuracy is possible with the use of the new Polyquant method that we have developed. Further work is however required particularly on the utility of this approach on a range of different phantoms, with different dose calculation algorithms, and also in particle therapy. EP-2070 Comparison of multi-atlas based synthetic CT generation methods for radiotherapy for prostate cancer C. Choi 1 , G. Sasso 2 , B. Pontre 1 1 The University of Auckland, Anatomy and Medical Imaging, Auckland, New Zealand ; 2 Auckland District Health Board, Radiation Oncology, Auckland, New Zealand Purpose or Objective Magnetic resonance imaging (MRI) plays an important role in radiation therapy planning (RTP) due to its excellent soft tissue contrast. In recent years, there has been increasing interest in MR-only RTP to prevent registration errors between the MRI and computed tomography (CT) images, to minimise total radiation dose to patients, and to reduce the time and cost associated with planning and treatment. The goal of this study is to evaluate the performance of MRI-guided synthetic CT (sCT) workflows used in a commercial RTP software package. Material and Methods A multi-atlas based sCT conversion workflow was developed using MIM’s in-built deformable registration tools (MIM Software Inc., Cleveland, OH). MIM offers three types of deformable registration: one based on image intensities (intensity-based), one using contours of anatomical structures of interest (contour-based), and one using a combination of the intensity-based and contour-based methods (hybrid). The contours for contour-based and hybrid workflows (bladder, prostate, rectum and bone) were manually drawn with reference to axial, sagittal, coronal MR images and validated by a clinician. The three workflows were tested on CT and T2 weighted MR atlases from six patients using a leave-one-out approach [1]. Once the atlas MR is registered to the target MR, sCTs were generated by applying the corresponding deformation vector field from the registration to atlas CTs, which were then fused to form a final sCT. Dice similarity coefficient (DSC) of the bone region and mean absolute error (MAE) of bone and tissue regions were computed in order to evaluate the performance of the three workflows. Results The target planning CT and sCTs produced from the three workflows are shown in Figure 1. Among the three deformable registration methods, the hybrid deformable registration workflow produced the highest dice coefficient, followed by the intensity-based and contour- based workflows (Table 1). Furthermore, the lowest mean absolute error was also produced using the hybrid workflow. However, the results show a reduction in grey scale contrast due to fusion of multiple sCTs. More focus should be on examining techniques to address this issue.

Figure 1: Graphical representation of dose results.