ESTRO 37 Abstract book

ESTRO 37

S526

breath hold; T1w 4D MRI was reconstructed using the joint MoCo-HDTV algorithm 1 . Dixon MRI was registered to the closest matching phase of T1w 4D MRI and transformed to MidP 2 using NiftyReg. First, pCT was generated from the MidP Dixon MRI using intensity-based segmentation and assignment of Hounsfield units (HU) for fat (-110), water (0), air (-1000) and lung (-767). Using machine learning (Gaussian process regression) on five sets of co-registered T1w and CT images, the relationship between MRI signal in lung and HU (-812 to 0) was established. Further, bony anatomy (HU 150 to 1500) was included by registering the best matching image from an atlas of fat and bone images (thresholded CT) to the Dixon fat images. Pseudo-CT with variable lung density and bone (bpCT) was then calculated by updating pCT of the first method using the machine learning and atlas method. MidP CT was generated from the 4D CT and registered to MidP pCT. One set of contours was generated for the primary tumour and thoracic organs at risk (OAR) after assimilating information from all MidP CT and MR images. VMAT plans at 6 MV were designed according to the RTOG 0617 standard arm (60 Gy in 30 fractions). Two planning techniques were used: initial planning on MidP CT and independent re-calculation on pCT and bpCT; and initial planning on MidP pCT (using both pCT methods independently) and re-calculating on MidP CT. Differences in dose-volume metrics were compared. Results Regarding OAR sites (Fig.1), the mean (over all patients and OARs) absolute differences in dose [cGy] relative to the original CT plan were only 9 ± 6 instead of 40 ± 32, and 9 ± 6 instead of 34 ± 27 for the original bpCT/pCT plans. The small dose differences for the OAR sites are likely to be clinically insignificant for the bpCT model. The largest mean (over patients) absolute differences in dose [cGy] corresponded to the planning target volume (PTV) D mean and were 46 ± 11 and 141 ± 27 when evaluating the CT plan on bpCT and pCT, and 45 ± 14 and 232 ± 189 for the original bpCT/pCT plans. The remaining minor differences of the bpCT model might be due to the heterogeneity of HU values in the tumour. Conclusion This study suggests that a clinically acceptable pCT model, with respect to OAR doses, can be calculated with a segmentation, machine learning and atlas approach. References: 1] Rank et al. Magn Reson Med . 2017. 2] Wolthaus et al. Med Phys . 2008.

Conclusion Posterior beam IMPT decreases dose to OARs compared to IMRT and at the same time increases the robustness towards anatomical changes during RT. Two fields can be used to reduce skin dose without loss of robustness. The highest robustness is achieved combining RO with PTV coverage, but an optimal strategy is lacking. PO-0959 Dosimetric Evaluation of Midposition Pseudo- CT for MR-only Lung Radiotherapy Treatment planning. J. Freedman 1,2 , H. Bainbridge 3 , A. Wetscherek 1 , D. Collins 2 , S. Nill 1 , A. Dunlop 1 , M. Kachelrieß 4 , M. Leach 2 , F. McDonald 3 , U. Oelfke 1 1 The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, Joint Department of Physics, London, United Kingdom 2 The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, CR UK Cancer Imaging Centre, London, United Kingdom 3 The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, Department of Radiotherapy, London, United Kingdom 4 German Cancer Research Center DKFZ, Medical Physics in Radiology, Heidelberg, Germany Purpose or Objective To perform a dosimetric comparison of lung radiotherapy treatment plans calculated on the midposition (MidP) image of a 4D pseudo-CT (pCT) with those calculated on a conventional MidP CT. The effect of including bone and variable lung density is studied. Material and Methods Four patients with locally advanced non-small cell lung cancer were scanned with a Philips Brilliance CT Big Bore scanner (Philips Medical Systems). MRI was obtained at 1.5 T (MAGNETOM Aera, Siemens Healthcare) using a radial T1w stack-of-stars spoiled gradient echo (GRE) sequence in free breathing and a 2-point Dixon GRE in