ESTRO 38 Abstract book

S547 ESTRO 38

registration uncertainties led to a median and 95 th percentile dose-CI of 0.0 Gy and 0.8 Gy, which were smaller than those of the difference between planned and accumulated dose: 1.1 Gy and 6.9 Gy respectively. The percentage of the voxels where the ΔD was outside of the dose-CI was 97.7 % (Figure 2).

To quantify local rigid registration accuracy, the first, third and last scans per patient were processed in a full circle method 1, The 4D reconstructions of these scans were made with an in-house developed coronal 4D-MRI sequence. This is based on retrospective image sorting of 30 repetitions of a multi-slice turbo-spin echo acquisition with an image resolution of 2x2x5 mm 3 . The full circle either consisted of the mid-position planning CT(pCT) and two 4D-MRI scans or alternatively, the pCT was replaced by the mid-ventilation of the first 4D-MRI scan. A shaped region of interest of the primary tumour and of each individual lymph node was used for local rigid registration in both procedures. From the values obtained from the full circle method in each patient, the mean and SD were divided by √3 to correct for the 3 registration steps in the circle. Patient characteristics were analyzed. Results 4D-MRI acquisition took about 4 minutes and showed adequate image quality (fig. 1) for 5 out of 6 patients. For the remaining 5 patients, all tumours were in the upper lobes and the average peak to peak amplitude of the primary tumour derived from the 4D-CT was 0.2, 0.4 and 0.2 cm in left/right, cranial/caudal and anterior/posterior respectively. The median GTV of the primary tumour was 88.3 cc and median GTV lymph nodes 49.9 cc. The corrected absolute mean of circle residuals was <0.11 cm for both primary tumor and lymph nodes in both methods. The corrected standard deviations are shown in table 1 ([1],2) .

Conclusion Small dose uncertainties were found. In the majority of voxels the uncertainties were smaller than the difference between planned and accumulated dose. This implies that DIR is sufficiently accurate to observe relevant changes in accumulated dose during treatments. PO-0994 Registration accuracy of 4D-MRI in lung acquired on the MR-linac M. Rossi 1 , M. Fast 1 , T. Van de Lndt 1 , M. Nowee 1 , J. Belderbos 1 , F. Lalezari 2 , J. Sonke 1 1 Netherlands Cancer Institute, Radiotherapy Department, Amsterdam, The Netherlands ; 2 Netherlands Cancer Institute, Radiology Department, Amsterdam, The Netherlands Purpose or Objective Differential motion between the primary tumor and mediastinal lymph nodes requires considerable safety margins in locally advanced lung cancer patients. The Unity MR-linac (Elekta AB, Stockholm, Sweden) provides high soft-tissue contrast and thus the potential to adapt the treatment plan to account for such differential motion. We recently have developed a self-sorting 4D-MRI method for daily IGRT of liver patients. The aim of this study was to determine the registration accuracy in 4D- MRI scans of NSCLC lung patients acquired on the MR- Linac. Material and Methods Six stage III NSCLC patients, receiving weekly 4D-MRI scans on the MR-linac were included in this study. Patients were treated on a conventional linac to a dose of 24x2.75 Gy.

Conclusion 4D-MRI registration of primary tumor and lymph nodes seems feasible on the MR-Linac with local rigid registration inaccuracies <0.2 cm facilitating daily online IGRT for locally advanced lung cancer patients. Registration accuracy potentially increases following further optimization of 4D-MRI contrast and image resolution tailored to lung cancer patients. 1 van Herk M et al. Med Phys 25:2054-2067, 1998 2 Schaake et al. IJROBP 2014 Nov 15 90(4) 959-66 PO-0995 An extension of van Herk’s margin recipe to explicitly account for time trends in tumor set-up. M. Giżyńska 1,2 , B. Heijmen 3 , P. Kukołowicz 1 1 Maria Skłodowska-Curie Memorial Cancer Center and Institute of Oncology, Medical Physics Department, Warsaw, Poland ; 2 University of Warsaw Faculty of Physics, Department of Biomedical Physics, Warsaw, Poland ; 3 Erasmus MC University Medical Center Rotterdam, Department of Oncology, Rotterdam, The Netherlands Purpose or Objective The most applied CTV-PTV margin recipe is that proposed by van Herk et al.: M =2.5 Ʃ + 0.7σ, where Ʃ and σ describe