ESTRO 38 Abstract book

S1144 ESTRO 38

1 University College London, Medical Physics and Biomedical Engineering, London, United Kingdom ; 2 Cambridge University Hospitals NHS Foundation Trust, Oncology, Cambridge, United Kingdom ; 3 University of Cambridge, Oncology, Cambridge, United Kingdom ; 4 NIHR University College London Hospitals Biomedical Research Centre, Medical Physics, London, United Kingdom Purpose or Objective A common problem with the QA of deformable image registration (DIR) algorithms is the absence of ground truths with which to generate quantitative metrics for performance evaluation. It was hypothesised that Effective Depth (ED) calculations could be used as a surrogate ground truth for assessment of image warping using DIR for daily proton dose evaluations. Material and Methods We compared two DIR algorithms, open-source software NiftyReg and commercial solution OnQ (Oncology Systems Ltd.), by deforming two head-and-neck CT scans to a repeated scan at the end of treatment for two different modalities: repeated CT (rCT) and image-guidance MVCT. Effective Depth (ED) Dose calculations were then performed using an in-house effective depth algorithm and stoichiometric calibration curves for the relevant modalities (MVCT and kVCT). The quality of the DIR performance was assessed according to APPM TG132 recommendations. By comparing the effective depth calculated, we can determine the range errors associated with the DIR that are relevant for dose calculations. Results We performed ED calculations for an equally-spaced range of 32 coplanar and non-coplanar beam angles centred on the GTV centroid as shown in Figure 1.

The results show that when the DIR is within tolerance according to the APPM recommendations, the %ED error is also within 3%. When the DIR is just outside the tolerance, as observed for the MVCT, the mean %ED error is between 3-4%. NiftyReg performed better than OnQ for both standard QA metrics and ED error for both modalities. The large %ED errors (>3%) for both NiftyReg and OnQ occurred at similar beam angles for both MVCT and rCT. However, the ED difference for NiftyReg and OnQ was systematically greater than the ground truth for MVCT but with opposite signs with respect to the ground truth for rCT. Conclusion This work highlights the potential for ED to be used as a QA measure for DIR for the purposes of proton dose calculations, which does not require labour-intensive manual contouring on the images involved. However, we also think this method would be useful for standard radiotherapy in centres using TomoTherapy with MVCT as daily IGRT. ED calculations could be used concurrently with contour propagation for QA of regular dose delivered assessments. Further work will be to implement this technique in a TomoTherapy centre and evaluating other commercial DIR software. A limiting factor to this work is that current proton centres use CBCT as their method of IGRT, which ED cannot be calculated on directly at present. EP-2075 mARC vs. IMRT prostate treatments: OAR dose distribution analysis stratified by PTV extent R. Bermúdez Luna 1 1 Hospital Universitario de Fuenlabrada, Medical Physics, Fuenlabrada, Spain Purpose or Objective mARC is the volumetric modulated arc therapy technique provided by Siemens. This technique was implemented in our centre during 2016. Among the reported benefits of the volumetric modulated arc therapy plans is that they can yield dose distributions highly conformed to the target volumes with improved protection of the surrounding organs at risk (OAR). The aim of this work has been to compare the dose distributions obtained with mARC plans with those corresponding to the IMRT technique in prostate cancer radiotherapy treatments. The analysis has been focused on the OARs and has been stratified according to seminal vesicle and pelvic lymph node involvement. Material and Methods The dose-volume histograms (DVH) of 60 prostate cancer mARC plans have been analysed and compared to the

The EDs were calculated for the deformed CT and the ground truth rCT and MVCT for comparison of the percentage difference. For the standard DIR QA, we calculated the mean surface distance, dice similarity coefficient (DSC) and Hausdorff distance (95 th percentile) for five organs at risk outlined on the CT, rCT and MVCT images: the parotid glands, submandibular glands and spinal cord. The results are shown in Table 1.