ESTRO 2025 - Abstract Book

S2996

Physics - Image acquisition and processing

ESTRO 2025

stopping-power ratio (SPR) prediction accuracy, using DECT and PCCT, to assess the suitability of CA scans for particle therapy treatment planning.

Material/Methods: This study utilized the Multi-Energy QA Phantom by QRM, containing tissue-equivalent inserts as well as CA inserts with different iodine concentrations between 0.5 and 15 mg/ml, distributed in solid water. Phantom scans were performed using two DECT (DECT1: SOMATOM go.Open Pro, DECT2: SOMATOM X.ceed) and one PCCT scanner (NAEOTOM Alpha) from Siemens Healthineers, with similar acquisition and reconstruction settings to ensure comparability. For all scanners, high X-ray tube voltages were selected to best penetrate the largest phantom setup, and an appropriate beam hardening correction (BHC) for iodine components was applied. RED and SPR maps were generated using the commercial CT manufacturer’s software. RED and SPR were extracted within defined ROIs (ROI diameter: 67% of insert diameter) and compared to the respective ground truth. The impact of varying CA concentrations and phantom sizes on RED and SPR prediction was evaluated (Fig. 1).

Results: For all phantom setups and CA concentrations, PCCT showed absolute ground-truth deviations in RED and SPR of less than 3%, while DECT exhibits higher deviations at high CA concentrations combined with larger phantom sizes (Fig. 2). Unlike the DECT scanners, SPR and RED predictions with PCCT were not influenced by the phantom size. Across all scanners, higher CA concentrations lead to increased RED deviation when comparing to the ground-truth, however this trend is less pronounced for PCCT. While DECT2 delivers near-constant RED and SPR predictions across CA concentrations and phantom setups, DECT1 exhibits notable deviations, especially at higher CA concentrations. The prediction of RED and SPR in tissue-equivalent inserts was not affected by the presence of CA (absolute SPR variation in cortical bone < 0.5%, not shown in Fig. 2).