ESTRO 2021 Abstract Book

S1387

ESTRO 2021

Conclusion The first step for implementing DD in clinical routine has been successfully accomplished as we can use one single calibration curve for all tube kV and standard deviation on HU remains within acceptable limits even for high electron density inserts. It would be necessary, however, to seek for dosimetric differences when comparing two plans, with and without DD algorithm.

PO-1665 Comparison of mid-position CT reconstruction systems M. Pereira 1 , J. Stroom 2 , F. Ghareeb 2 , D. Boukerroui 3 , M. J. Gooding 3 , C. Greco 2

1 Faculdade de Ciências, Universidade de Lisboa, 1749-016 Lisboa, Portugal; 2 Champalimaud Centre for the Unknown, Department of Radiation Oncology, Lisbon, Portugal; 3 Mirada Medical Ltd., Science, Oxford, United Kingdom Purpose or Objective The application of the mid-position CT scan (MPCT) for cancer patients with respiratory moving targets should lead to smaller treatment volumes, potentially boosting the delivery of hypofractionated treatments. We aim to assess the accuracy of our in-house freeware-based solution to generate MPCT, and comparing it to an automated research prototype being developed by Mirada Medical. Materials and Methods 4DCT data from 3 digital phantoms and 18 cancer patients (8 lung, 6 liver and 4 pancreas) was used in this study. Two of the digital phantoms are based on the noise-free 4D extended cardiac-torso (XCAT) phantom, in which organs have predefined voxel values and motion. A well-defined spherical lesion (1.0 and 2.0 cm) was placed near the diaphragm. The third digital phantom was in-house generated, and provided us with ground truth data. In the open-source medical imaging software 3DSlicer, the Elastix registration toolbox was used to register the 10 breathing phases of the 4DCT. The resulting deformation vector fields (DVF) were used to generate the MPCT. DVFs were also employed to calculate the target’s (CTV, or implanted marker for liver and pancreatic patients) motion amplitude (MA). The MAs were compared to our reference amplitudes, from manual determination of the center of mass (COM) in all breathing phases. The reference mean position (MP) over the respiratory cycle of the target was compared to its position in MPCT reconstructed by both softwares. The MPCTs were compared in terms of image difference and image quality comparing them to the acquired planning CT (pCT) by visual inspection. Results Table 1 depicts the difference between the reference and the calculated target’s MP and MAs. The differences regarding MP and MAs are within 1 mm, thus showing an accurate patient's anatomy MP reconstruction and MAs calculation for both softwares. The mean MPCT image difference was about 1 HU for all the analyzed cases, but it was bigger for the digital phantoms (LP1 and LP2), possibly because registration benefits from images having some texture. Image registration might deviate in homogeneous regions, where we observed higher HU differences, particularly for liver and pancreatic cancer patients (Fig.1). These HU differences will not produce meaningful differences in dose calculation. Although MPCTs reduce image artefacts due to breathing, yielding smaller CTV-PTV margins, their reconstruction resulted in some image blurring (Fig.1), possibly due to residual motion or change of anatomy in respiratory phases of 4DCT.