ESTRO 2025 - Abstract Book

S3020

Physics - Image acquisition and processing

ESTRO 2025

2575

Digital Poster Longitudinal study on synthetic CT constancy and reproducibility: Assessing QA feasibility with volunteer scans Bernd-Niklas Axer 1,2,3 , Florian Putz 1,2,4 , Rainer Fietkau 1,2,4 , Christoph Bert 1,2,4 , Sultan Haider 3 , Juliane Szkitsak 1,2,4 1 Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen Nürnberg (FAU), Erlangen, Germany. 2 Comprehensive Cancer Center Erlangen-EMN, CCC ER-EMN, Erlangen, Germany. 3 Innovation Think Tank, Siemens Healthineers AG, Forchheim, Germany. 4 The Bavarian Cancer Research Center, BZKF, Erlangen, Germany Purpose/Objective: Registration errors between CT and MR images lead to uncertainties in radiation treatment planning of brain tumors [1]. Using an MR-only workflow and generating synthetic CT (sCT) can prevent this issue. However, a standardized quality assurance (QA) protocol for MR-only workflows has not yet been established [2]. To assess the feasibility of QA based on repetitive volunteer scans, a longitudinal study is conducted with healthy participants to evaluate the reproducibility and constancy of the generated sCTs. Material/Methods: Each volunteer undergoes a weekly series of three identical T1-VIBE-Dixon-sequences (slice thickness: 0.8mm) in irradiation position with thermoplastic full-face-mask (Brainlab AG). Body weight, head and neck circumference are measured beforehand. The participant remains immobilized between the first two sequences, followed by full repositioning before the third. Afterward, volunteers evaluate immobilization via a questionnaire. Five sCTs are generated from each T1-sequence using a commercially available algorithm (syngo.via RT Image Suite (VB60), Siemens Healthineers AG). To date, 22 series have been acquired from eight participants, ranging from one to five series each, resulting in 22x3x5=330 sCTs. After rigid registration, these sCTs are analyzed using adapted global (γ GLO ) HounsfieldUnit-γ-PassRates (HUγPR). The γ-criteria used are γ 1 =1%/1mm, γ 3 =3%/3mm and γ 5 =5%/5mm. The following combinations of two sCTs are analyzed: based on single T1-sequence (A), without repositioning within one series (B), with repositioning within one series (C), and with a minimum five-day gap between two series, using the first T1-sequence from each (D). Results: For combination A, a mean HUγPR A =100.0±0.0% with γ 1 is observed. Combination B shows mean HUγPR B of 97.3±2.2%, 99.7±0.4% and 99.9±0.1% with γ 1 , γ 3 and γ 5 , respectively. Combination C provides mean HUγPR C values of 90.3±3.8%, 97.0±2.2% and 98.0±2.1%. Combination D shows mean HUγPR D of 87.6±3.1%, 96.0±1.7% and 97.4±1.6%.