This retrospective, single-center study was approved by our Institutional Review Board. The requirement for informed consent was waived.

Patients

Based on earlier studies [8, 11], our institution adopted a low-iodine-dose protocol for pediatric abdominal CT scans in April 2022. In July 2022, we retrospectively reviewed the picture archiving and communication system (PACS) (Infinitt Healthcare, Seoul, South Korea) database for the period from January 2022 to July 2022 and searched for sequential abdominal CT scans—those with a standard iodine dose protocol (standard-dose group) and those with a low-iodine-dose protocol (low-dose group)—within a 4-month interval for the same patient. The standard-dose group were given Iobitridol 350 mgI/mL (Xenetix 350, Guerbet, Aulnay, France) with an iodine dose of 595 mgI/kg, while the low-dose group were given Iohexol 240 mgI/mL (Iobx 240, Taejoon Pharmaceutical Co., Seoul, Korea) with an iodine dose of 408 mgI/kg. Of the 46 potential participants, the following were excluded: (1) patients aged ≥ 18 years (n = 4), (2) patients whose body weight changed by more than 5% between the two scans (n = 5), and (3) images displaying severe motion or respiratory artifacts that impeded image interpretation (n = 2). Subsequently, a total of 35 pediatric patients were included in this study.

Computed tomography protocols

A 128-channel multidetector CT (MDCT) scanner (SOMATOM Definition Flash, Siemens AG, Forchheim, Germany) was used to obtain the CT images. In total, 1.7 mL/kg of contrast agent was administered over 50 s using a power injector (Envision CT, Medrad, Pittsburgh, PA). The average contrast agent volume was 55.2 ± 28.9 mL for the standard-dose protocol and 55.5 ± 28.5 mL for the low-dose protocol. The average injection rate was 1.1 ± 0.6 mL/s. The portal venous phase was obtained 70 s after contrast injection. DECT protocol was employed utilizing two X-ray tubes operating at different tube voltages—70 kV and 150 kV—with reference tube current–time products set at 370 mAs for 70 kVp and 93 mAs for 150 kV. Detailed CT scan parameters and contrast media usage for both standard- and low-dose CT are presented in Table 1.

Table 1 Computed tomography scan parameters and contrast media for each protocolArtificial intelligence-based contrast-boosting technique

For the AI-based contrast-boosting technique, ClariACE (ClariPi, Seoul, South Korea) was used. ClariACE has a two-stage U-net architecture to enhance contrast in low-contrast-dose CT. The details of ClariACE are described in Supplementary Material 1.

Quantitative image analysis

Quantitative analysis was conducted by a third-year radiology resident (D-J.S.) who was blinded to clinical information. CT images were assessed in soft tissue window settings (width: 400 HU, level: 50 HU). The CT Hounsfield unit (HU) was measured on the axial image where the portal vein was observed at full width. Attenuations of liver parenchyma, portal vein, aorta, paraspinal muscle, and extracorporeal air were measured by manually drawing regions of interest (ROI) based on visual inspection. The ROIs were drawn at sites with homogeneous attenuation, and the size and shape of each ROI were applied equally to each organ. Regions with inhomogeneous attenuation due to vascular structures or beam-hardening artifacts were excluded from measurement. The average HU value of four separate ROIs in the right anterior, right posterior, left medial, and left lateral segments represented the attenuation of liver parenchyma. Similarly, the average HU value of two ROIs on both sides of the paraspinal muscles represented muscle attenuation. Image noise was determined as the average of two standard deviation (SD) values from both paraspinal muscles. The contrast-to-noise ratio (CNR) and the signal-to-noise ratio (SNR) were calculated as CNR = \((_-_)\)/\(_\) and SNR = \(_\)/\(_\), where \(_\) is the HU value of paraspinal muscle.

Qualitative image analysis

Qualitative image analysis was conducted independently by D-J.S. and Y.H.C. (a pediatric radiologist with 17 years of experience), blinded to the clinical information. The readers evaluated organ enhancement, vessel enhancement, noise, artifact, overall image quality, and lesion conspicuity of the CT images, on a 5-point scoring scale. The grading scales for subjective image quality items are summarized in Table 2. Organ enhancement, vessel enhancement, overall image quality, and lesion conspicuity were scored higher in better image quality. Higher noise and artifact scores indicated less noise and minimal artifacts. Before assessing lesion conspicuity, all lesions were annotated on the PACS by a second pediatric radiologist (S.B.L. with 7 years of experience). The average qualitative scores from both readers were used for statistical analysis.

Table 2 Grading scale for subjective image qualityRadiation dose and iodine uptake

The CT dose index volume (CTDIvol, mGy) and dose-length products (DLP, mGy × cm) were recorded for all CT examinations based on CT dose reports. The effective dose (ED, mSv) was calculated as ED = DLP × K, where K is the conversion factor for the abdomen, which varies depending on kVP and age [20]. As the conversion factors for less than 80 kVp are unknown, a conversion factor of 80 kVp was applied instead. Linear interpolation was used to calculate the conversion factors for different ages and kVp values. The total iodine uptake for each patient was calculated as body weight (kg) × 1.7 mL/kg × iodine concentration (mgI/mL).

Statistical analysis

Continuous variables were summarized as means and SD. Categorical variables were summarized as counts. For quantitative analysis, repeated measures one-way analysis of variance (ANOVA) with post hoc pairwise comparisons with Bonferroni correction was used to compare the three groups: standard-dose group, low-dose group, and low-dose group with AI-based contrast-boosting (contrast-boosted group). Qualitative scores were analyzed using the Friedman test, followed by post hoc Dunn’s pairwise comparisons. Interobserver agreement was assessed using the intraclass correlation coefficient (ICC): poor (< 0.20), fair (0.21–0.40), moderate (0.41–0.60), good (0.61–0.80), or excellent (0.81–1.00) [21]. The paired t-test was used to compare radiation dose and iodine uptake between the standard- and low-dose groups. All analyses were performed using MedCalc Software (version 22.009, MedCalc Software Ltd, Ostend, Belgium). Statistical significance was set at P < 0.05.