ESTRO 2025 - Abstract Book

S3005

Physics - Image acquisition and processing

ESTRO 2025

2156

Digital Poster FIRST IMPLEMENTATION OF PHOTON COUNTING CT FOR OPTIMIZING SEGMENTATION AND RADIOTHERAPY PLANNING FOR HEAD-AND-NECK CANCER Niccolo Bertini 1,2 , Hubert S Gabrys 1 , Hatem Alkadhi 3 , Lotte Wilke 1 , Patrick Wohlfahrt 4 , Serena Psoroulas 1 , Eugenia Vlaskou 1 , Laura Motisi 1 , Matthias Guckenberger 1 , Stephanie Tanadini-Lang 1 , Panagiotis Balermpas 1 1 Department of Radiation Oncology, University Hospital Zurich, University of Zurich, Zurich, Switzerland. 2 Radiation Oncology Unit, Oncology Department, Azienda Ospedaliero Universitaria Careggi, Florence, Italy. 3 Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Zurich, Switzerland. 4 Siemens Healthineers, Varian, Cancer Therapy Imaging, Forchheim, Germany Purpose/Objective: Photon-counting detector computed tomography (PCD-CT) offers advancements over conventional energy integrating detector CT (EID-CT) by counting individual X-ray photons and capturing energy resolved information. This study is the first to investigate the use of PCD-CT for delineation purposes, comparing the accuracy of organ and target segmentation between PCD-CT and EID-CT systems. Material/Methods: Twelve patients with head-and-neck cancer and target volumes in close proximity to radio-opaque oral implants were prospectively included, with imaging data for all patients collected twice, during routine radiotherapy-planning (EID-CT) and diagnostic (PCD-CT) procedures. Delineation of the clinical target volume (CTV) and organs at risk (OARs) was performed by three physicians and an AI-based system (MIM Contour ProtegeAI) using EID-CT (120 kVp) and PCD-CT at virtual monoenergetic levels of 140 keV (for reducing metal artifacts) and 50 keV (for optimizing the iodine contrast) ( figure 1 ). Segmentation accuracy was assessed using the Dice similarity coefficient (DSC). Figure 1: Same slice from two reconstructions of a PCD-CT scan from the study, left: 50 keV VMI, right: 140 keV VMI.

Results: The ratio of the mean computed tomography dose index (CTDI) for EID-CT/PCD-CT for all patients was 2.72. The study revealed high inter-observer agreement for the mandible, oral cavity, parotid glands, and submandibular glands, with DSCs exceeding 0.8 across both EID-CT and PCD-CT. No significant differences in delineation agreement were observed for most regions of interest, while a significant improvement in agreement was found for the lips (p = 0.001) and oral cavity (p = 0.012) when combining 140 keV and 50 keV images in PCD-CT compared to EID-CT ( figure 2 ). When comparing human experts with the AI model, high agreement was maintained for most structures across both CT types.