This retrospective study was conducted in accordance with the Helsinki declaration and approved by the Institutional Review Board and Ethics Committee of the Hallym University Sacred Heart Hospital (IRB number: 2021-08-024-001). The requirement for written informed consent was waived and confirmed by the Institutional Review Board and Ethics Committee of the Hallym University Sacred Heart Hospital. This study design followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
1. Sample size estimation and patient enrollment
This study was designed for matched pairwise comparison. The minimum sample size was determined to 44 patients under the assumption of effect size of 0.5, a priory statistical significance of 0.05 and a power of 90%.
From July 2020 to June 2021, 371 patients received 90kVp tube voltage and reduced contrast agent administration protocol. Among them, 57 patients above 18 years of age received both 90kVp and 100kVp tube voltage PDCT within the same period. Six patients were excluded because of severe motion artifacts (n = 1), severe generalized edema-related pancreatitis or recent pancreatic surgery (n = 3), and no information on body mass index (n = 4) (Figure 1). Ultimately, 51 patients were enrolled in the study.
2. CT protocol
All CT images were obtained using one of three machines. The reduction protocol group was scanned with 90kVp tube voltage and 374 mAs or 393 mAs of reference tube current using MDCT scanners (SOMATOM Force, Siemens Healthineers) with a reduction of the standard amount of contrast material (300 mg of iodine per mL, approximately 1.4 ml/kg, according to the dosage adjustment table (Table 1) and reconstructed by advanced IR algorithm (ADMIRE strength 2, Br40 of the reconstruction kernel). The standard protocol group was scanned with a 100kVp tube voltage and 289 mAs of reference tube current using two MDCT scanners (SOMATOM Definition Edge or SOMATOM Definition Flash; Siemens Healthineers) using a standard amount of contrast material (300 mg of iodine per mL, 2 ml/kg) and reconstructed using the IR algorithm (SAFIRE strength 2, I40 of reconstruction kernel). PDCT consists of a dual phase (arterial and portal venous phases). All patients received an intravenous nonionic contrast medium containing an iodine concentration of 300 mg/mL (Bonorexâ 300 [iohexol], Central Medical Service, Korea; Iomerolâ 300 [iomeprol], Bracco Imaging, Korea). For dynamic imaging, nonionic contrast material per kilogram of body weight was administered at a rate of 25 mL/s using an automatic power injector (Multilevel CT, Medrad, USA), followed by a 20 mL flush of sterile saline. Bolus tracking method was used to determine the timing of arterial phase scanning, and the arterial phase was obtained 15 seconds after triggering when the proximal abdominal aorta reached the Hounsfield unit of 100 or greater. Portal venous phase images were acquired 30 seconds after the end of arterial phase scan. The reconstruction parameters of the axial image were a 3 mm section thickness and a 3 mm reconstruction interval. A coronal reformatted image was reconstructed with a section thickness of 3 mm and an interval of 3 mm. Table 2 is comparison of both CT acquisition protocols.
3. Radiation dose evaluation
The radiation dose of the pre-contrast images was excluded from the evaluation because the kVp of the pre-contrast scan was not fixed. The only radiation dose in the portal phase images was compared. Size-specific dose estimation was calculated using the sum of the anteroposterior and transverse dimensions according to the report 204 released by the American Association of Physicists in Medicine (AAPM) . The effective tube current (mAs) of each patient in the portal phase was recorded.
4. Image analysis
4.1 Objective assessment of image quality
The standard deviation (SD) of the air located outside the patient's xyphoid process level was defined as the objective image noise. The mean CT Hounsfield units (HU) of each abdominal organ, such as the aorta, liver, main portal vein (MPV), pancreas, spleen, kidney, and psoas muscle, as well as cystic lesions of the pancreas and liver which is more than 5 mm in diameter, were measured using a circular region of interest. When drawing ROIs in each target organ, special care was taken not to include adjacent vessels, bile ducts, artefacts, or peritoneal fat.
The mean CT HU of the aorta, pancreatic parenchyma, and psoas muscle were measured in both the arterial and portal phases. The mean CT HU values of the liver, MPV, spleen, and kidney were measured only during the portal phase of CT scanning. Liver attenuation was recorded as the mean of the measurements of four ROIs in the medial and lateral segments of the left hepatic lobe, and the anterior and posterior segments of the right hepatic lobe . Aortal attenuation was measured at the celiac trunk take-off level. The attenuation of the psoas muscle was recorded as the mean attenuation of two ROIs that avoided macroscopic fat infiltration at the L4 vertebral level. Kidney attenuation was measured in the renal cortex, with special care taken to avoid containing the medulla and perirenal fat. The size, shape, and position of all ROI measurements were kept constant by applying a copy-and-paste function at the workstation. The SNR and CNR were calculated as follows:
SNRtarget organ = HUtarget organ / background image noise
CNRtarget organ = (HUtarget organ – HUpsoas muscle) / background image noise
4.2. Subjective image quality analysis
The subjective image quality analysis was independently and blindly evaluated by two board-certified radiologists with more than 10 years of experience. One reviewer was from an off-site institution. Subjective image noise, beam hardening or streak artefacts, visibility of small structures (peripheral hepatic vessels), lesion conspicuity, and overall diagnostic confidence were evaluated using a 5point scale based on the European Guidelines on Quality Criteria for Computed Tomography and previous research published in the radiology literature [7, 24-26]. All 102 PDCT sets were reviewed by two reviewers without any information of patient and scan technique. Subjective image noise was graded on a 5point scale based on the presence and amount of image mottle or graininess (5, minimal image noise; 4, less than average noise; 3, average image noise; 2, above average noise; 1, unacceptable image noise). The visibility of small structures, mainly hepatic vessels, was also graded using a 5-point scale, with 5 indicating excellent visualization and 1 indicating imperceptible small hepatic vessel structures. Beam hardening or streak artefacts were graded on a 5point scale (5, absence of artefact; 4, mild artefacts not interfering with diagnosis; 3, moderate artefacts slightly interfering with diagnosis; 2, pronounced artefacts interfering with diagnosis; and 1, impossible interpretation of a lesion or an organ of interest).
Hepatic cysts, pancreatic cystic lesions and hepatic hemangiomas more than 5mm in short diameter were selected by an independent researcher for lesion conspicuity evaluation. Each reviewer received brief lesion information including image number on arterial or portal phase, anatomic location or adjacent anatomic landmark.
Lesion conspicuity was ranked on a 5point scale with a score of 5 indicating a clearly seen lesion with clearly visualized margins and a score of 1 indicating an imperceptible lesion. Lesion conspicuity was evaluated based on the visibility of the lesion boundary. When more than 75% of the boundary of the lesion was visible, it was marked as 5 points; when 50%75% of the boundary was visible, it is marked as 4 points; when 2550% of the boundary was visible, it is marked as 3 points; and when only less than 25% of the boundary was visible, it was marked as 2. Overall diagnostic confidence was evaluated using a 5point scale; grade 1, nonconfident; grade 2, subdiagnostic confidence; grade 3, average confidence; grade 4, more than average; and grade 5, completely confident.
Continuous variables are expressed as means and SD. Intraobserver and interobserver agreements were assessed using the weighted kappa statistics. Quantitative image parameters (attenuation values, image noise, SNR, and CNR) were corrected using Welch's test depending on normality testing and compared using paired t-tests. Qualitative subjective image analysis of the image was performed using the Wilcoxon signed-rank test. Statistical analysis was performed using the MedCalc software (MedCalc 13.1.2). Statistical significance was set at p < 0.05.
The dataset used in this study is not publicly available due to patient privacy constraints and data availability policy of our institution committee. An anonymized version of this dataset can be available from the corresponding author on reasonable request.