ESTRO 2025 - Abstract Book

S3064

Physics - Image acquisition and processing

ESTRO 2025

3789

Digital Poster Quantitative and qualitative assesment of a novel algorithm for correcting stack-artifacts in 4DCT Reconstruction Mushawar Ahmad 1,2,3 , Maximilian Heider 1,3,4 , Niklas Lackner 1,3,4 , Oliver J Ott 1,3,4 , Rainer Fietkau 1,3,4 , Jannis Dickmann 2 , Christoph Bert 1,3,4 , Juliane Szkitsak 1,3,4 1 Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen Nürnberg (FAU), Erlangen, Germany. 2 Cancer Therapy, Siemens Healthineers AG, Forchheim, Germany. 3 Comprehensive Cancer Center Erlangen-EMN, CCC ER-EMN, Erlangen, Germany. 4 The Bavarian Cancer Research Center, BZKF, Erlangen, Germany Purpose/Objective: Irregular breathing amplitudes during 4DCT data acquisition cause motion artifacts, particularly stack-misalignments, compromising image quality [1]. This study aimed to evaluate image quality and anatomical correctness of a new prototype stack-artifact correction reconstruction algorithm. Material/Methods: A retrospective study was conducted comparing ZeeFree (Siemens Healthineers AG,Forchheim,Germany) against standard reconstructions. ZeeFree is based on a DEMON joint 3D vector field algorithm to correct stack misalignments. CT data was acquired using the i4DCT algorithm [2] as part of the clinical routine and reconstructed for both reconstruction algorithms as well as amplitude- and phase-based binning. The study included 20 patients with tumors in the lung with high amplitude variability and 10 randomly chosen patients for verification. Quality of 30x2x2=120 CT datasets was rated by three radio-oncologists and three medical physicists. Ratings were given on a Likert scale ranging from 1 (severe artifacts) to 5 (no artifacts) and anatomical correctness was assessed using a binary evaluation. Statistical significance was assessed using Cohen’s-d, and the strength of association between variables was evaluated with Chi-Square tests and Cramér’s-V. Additionally, organs at risk (OARs) were segmented using the TotalSegmentator [3] and analysed to quantify differences in the CT values, volumes, and Dice similarity scores.

Results: ZeeFree reconstructions showed considerable improvements over Standard. Less than 10% of images with ZeeFree were rated 1 or 2, compared to 30% with Standard, while 75% of ZeeFree images were rated 4 or 5, compared to 40% with Standard. Additionally, 40.6% of cases (73/180) improved from a rating of 3 or lower to 4 or 5 with ZeeFree. For anatomical correctness, 82.4% (61/74) of cases classified as anatomically incorrect with Standard were rated anatomical correct with ZeeFree, while for 2.8% (3/106) of cases rated as correct with Standard the rating decreased to incorrect with ZeeFree.