ESTRO 2025 - Abstract Book

S2949

Physics - Image acquisition and processing

ESTRO 2025

study presents the implementation of synchronized contrast-enhanced 4DCT (ce4DCT) simulation to improve target delineation and respiratory motion management in abdominal SBRT planning.

Material/Methods: Twenty-two patients were simulated with two sequential 4DCT scans: one at baseline and one contrast-enhanced with personalized delay-time ( t delay ) calculated to capture the tumor in the desired contrast phase, based on previous diagnostic triple-phase CT. The internal target volume (ITV) was delineated on the ten contrast phases, rigidly registered to the baseline scan. Aortic HU values were measured over time on the ce4DCT to refine t delay for subsequent patients. The standard approach, combining triple-phase CT with unenhanced 4DCT, was simulated. The differences in target delineation were evaluated by volume, centroid shift, Dice and Jaccard indices, and mean distance agreement (MDA). Margins that would have been required to account for target motion were calculated. Results: All ce4DCT scans provided clear delineation of anatomical structures and vessels, significantly improving tumor visibility over the entire breathing cycle in 20 out of 22 cases. The median contrast peak time was 54.5 seconds and washout plateau at 70.3 seconds, with mean peak and plateau HU values of 292 ± 59 and 169 ± 25, respectively. Volumes contoured using the standard procedure (ITV2) were significantly smaller than the ones from our approach (ITV1) (P=0.045). The median centroid shift was 4.7 mm (range, 0.9 – 12.6). ITV1-ITV2 overlap was 69% (Dice index), 53% (Jaccard index), and 2.89 mm (MDA), with liver volumes showing lower indices compared to pancreatic volumes. Margins required to better encompass ITV1 were highly variable, with liver tumors requiring an average of ≥5 mm in most directions, while pancreatic tumors needed over 3 mm only anteriorly and inferiorly. Conclusion: The ce4DCT simulation was feasible, improved target delineation with minimal resource investment, and reduced uncertainties in SBRT planning by addressing poor tumor visibility and respiratory motion challenges. The combination of triple-phase 3DCT and unenhanced 4DCT introduces great variability in target delineation, making isotropic margins ineffective in accurately assessing the target motion without unnecessarily including surrounding tissues. Digital Poster Rapid Anatomical Verification for Prostate Cancer Patients Using Deep Learning-Generated Volumetric Images from Biplanar Projections and Planning CT Mustafa Kadhim 1,2 , Emilia Persson 2,3 , André Haraldsson 1,2 , Mikael Nilsson 4 , Christian Jamtheim Gustafsson 2,3 , Malin Kügele 1,2 , Sven Bäck 1,2 , Sofie Ceberg 1 1 Medical Radiation Physics, Lund University, Lund, Sweden. 2 Hematology, Oncology, and Radiation Physics, Skåne University Hospital, Lund, Sweden. 3 Translational Medicine, Lund University, Malmö, Sweden. 4 Centre for Mathematical Sciences, Lund University, Lund, Sweden Purpose/Objective: Precise localization of tumors and surrounding anatomy is critical to enable high precision radiotherapy of prostate cancer. While daily biplanar X-ray imaging can be used for rapid verification of patient positioning, these 2D-images are insufficient to fully capture the 3D anatomical changes occurring during treatment. Consequently, cone-beam CT (CBCT) might be utilized as a pre-treatment imaging modality. Nonetheless, relying on CBCT-images may increase treatment timeslot and offer higher radiation exposure to the patients. In this work, we propose a novel deep learning (DL) approach that combines anatomical information from both daily biplanar images and treatment Keywords: Contrast imaging, abdominal SBRT, breathing motion 492