Application of AIIR algorithm for quality improvement and noise reduction in pediatric abdominal contrast-enhanced CT

To investigate the feasibility and value of the deep learning full model iterative algorithm (AIIR) in reducing radiation dose during contrast-enhanced whole abdominal CT scans in children.

Data from 100 pediatric patients undergoing contrast-enhanced whole abdominal CT scans due to clinical indications were retrospectively collected. The patients were divided into a conventional group (n = 50,100 kVp, automatic current modulation 100 mAs) using hybrid iterative reconstruction (HIR) and an experimental group (n = 50,80 kVp, automatic current modulation 180 mAs) employing HIR and AIIR reconstruction. Subjective image quality scores (five-point scale evaluation), signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of major organs, contrast agent dosage, and radiation dose (CTDIvol, DLP, ED) were compared among the three groups during the arterial phase.

The experimental group exhibited significantly lower radiation dose compared to the conventional group (ED reduction by 23.61%, P < 0.001). In terms of image quality, subjective scores indicated superior performance of the AIIR group over both conventional HIR and experimental HIR groups (P < 0.05), with excellent inter-physician agreement (Kappa>0.70). The AIIR group demonstrated significantly higher SNR and CNR values for the liver, spleen, kidneys, gallbladder, pancreas, and abdominal aorta compared to the HIR groups (all P < 0.05).

The AIIR algorithm achieves superior image quality at low dose levels (80 kVp) compared to conventional HIR and low-dose HIR protocols, demonstrating significant clinical value and potential for widespread application in pediatric abdominal imaging.