ESTRO 2023 - Abstract Book

S557

Sunday 14 May 2023

ESTRO 2023

period of 4 seconds, oriented in cranial/caudal (CC) direction. Three types of scans were tested (T2_3D_TSE: SENSE factor 3; T2_3D_TSE_CS: CS factor 5; T1_3D_VANE XD Dixon: water image) with a reconstructed voxel of 0.64x0.64x2/0.64x0.64x2/0.78x0.78x1.5mm3 and a scan duration of 4:04/2:52/5:43 minutes, respectively. The spherical target was delineated. From the delineations the cranial/caudal (CC) length was extracted, which is a surrogate to measure if the GTV (30 mm), the ITV (30+20 = 50 mm) or a volume in between is shown. In addition, the center-of-volume (CoV) of the delineation was determined. A high and low contrast imprint of the sphere was visible in all motion images therefore a full (automatic using threshold) and tight (most prominent imprint of the spherical sphere only) delineation was performed (figure 1).

Results Static scans confirmed a CC sphere diameter of 30 mm for all 3 imaging techniques. For the motion scans, this full delineation diameter increased to 48/42/46.5 mm for T2 TSE/T2 TSE CS/VANE respectively (figure 2). The CoV deviated - 1/0/-1 mm from the time-averaged (mid-)position. For the tight delineation a CC diameter of 30/36/28.5 mm was observed, while the CoV shifted to 5/1/7 mm compared to mid-position .

Conclusion The observed CC diameter is largest in T2_3D_TSE and T1_3D_VANE and coincides with the definition of the ITV, while for T2_3D_TSE_CS a markedly decreased dimension was observed. CoV positions were close to mid-position for all techniques. For the tight delineation, the CC diameter of T2_3D_TSE and T1_3D_VANE were close to the actual sphere , however at the largest deviations from mid-position; especially T1_3D_VANE is positioned closer to end-expiration position. This end expiration position is the most prevalent position (but not the time average position) in the cos4 signal which therefore prevailed in the T1_3D_VANE reconstruction. For CS reconstruction (T2_3D_TSE_CS), motion artifacts seem to be partly suppressed by the iterative reconstruction process which removes the non-coherent artifacts arising from sparse sampling. Alterations in volume and position due to MRI acquisition/reconstruction techniques need to be considered in treatment planning and delivery. 1 The University of Manchester, Division of Cancer Sciences – Faculty of Biology, Medicine and Health , Manchester, United Kingdom; 2 European Bioinformatics Institute, European Molecular Biology Laboratory, Hinxton, United Kingdom; 3 The Christie NHS Foundation Trust, Christie Medical Physics and Engineering, Manchester, United Kingdom Purpose or Objective Cycle-consistent generative adversarial networks (cycleGANs) are used to generate synthetic CTs (sCTs) from on-treatment CBCTs for radiotherapy treatment verification and enable daily plan adaptation. For optimal usability in the clinic, image synthesis models must be generalisable between systems and centres. Unlike CT machines, XVI systems do not give electron density information needed for a dose calculation, and CBCT image quality parameters are often machine-specific, despite the same machine model, software, and parameters functioning within manufacturer tolerance. We investigate the impact of training with CBCT images from single vs. multiple systems for sCT generation to determine whether capturing such variability within training protocol leads to more generalisable models. PD-0665 Do CBCT-based image synthesis models need to be machine-specific? C. Sargeant 1 , J. Shortall 1 , A. Green 2 , R. Chuter 3,1 , A. McWilliam 1

Materials and Methods