Computed Tomography System Performance for Different Iterative Reconstruction Algorithms

Abstract

The number of the Computed Tomography (CT) examinations has increased greatly over the past decade; This has made CT doses the largest source of the individual effective dose from medical exposures worldwide. Evaluation of image quality in CT is an important issue to ensure diagnostic accuracy, while keeping the radiation dose especially to the pediatric patient as low as reasonably possible. The purpose of this study was to assess the physical and psychophysical image quality in pediatric CT for different iterative reconstruction (IR) algorithms. The image quality was measured on each CT system using pediatric QRM abdomen and thorax phantoms for clinical thorax and abdomen examination scanning parameters and using the Catphan 600 phantom for head examination scanning parameters. The performance of each CT systems was evaluated individually using conventional filtered back-projection (FBP) and IR algorithm employed in that system. Noise, contrast-to-noise ratio (CNR), inverse image quality figure (IQFinv), modulation transfer function (MTF) and noise power spectrum (NPS) were assessed for FBP and wide range of iteration level of different IR algorithms. Figure of Merit (FOM) parameter was calculated to characterize the dose efficiency of each CT system. The CTDIvol values were changed from 20.6 to 45.7, 0.9 to 7.4 and 2.4–9.6 mGy for pediatric head, thorax and abdomen scanning protocols between CT systems, respectively. With increasing iteration level, image noise decreased while CNR and IQFinv increased. NPS peak was decreased with increasing iteration level for both two (2D) and three dimensional (3D) NPS and the ratio of the 3D NPS to 2D NPS varied from 4.5 to 8.3 for pediatric head, from 4.6 to 8.7 for pediatric thorax and from 4.0 to 14.0 for pediatric abdomen scanning protocols. For all IR algorithms employed with CT systems used in this study, the FOM values were increased with increasing iteration level. © 2024 Elsevier Ltd