This single-center retrospective study was conducted at a university medical hospital and was approved by the institutional review board and local ethics committee. The requirement for informed consent was waived because this was a retrospective, non-interventional, observational study. Consecutive patients who underwent multiphasic abdominal PCD-CT including the early arterial phase were enrolled between March and April 2023. The exclusion criteria were patients who had undergone abdominal vascular surgery were excluded (Fig. 1).

Flowchart of patient enrollment

All CT images were acquired using a clinical dual-source PCD-CT scanner (NAEOTOM Alpha, Siemens Healthineers, Forchheim, Germany). All scans were acquired in single-source, multi-energy mode (Quantum Plus) using a tube voltage of 120 kV, collimation of 120 × 0.2 mm, pitch factor of 0.8, and gantry rotation time of 0.50 s. The VMI energy levels were 50 and 70-keV. The low-keV energy was set to 50-keV because it has been reported that 50-keV VMI has the best image quality in PCD-CT for abdominal CTA [23, 24]. As recommended by the manufacturer, the tube current–time product was set to an image quality level of 170, and automatic tube current modulation (CARE Dose4D; Siemens Healthineers) was used. The protocol for upper abdominal multiphasic CT imaging was 600 mgI per unit body weight (kg) of the nonionic iodine contrast material, iomeprol (Iomeron 350; Eisai, Tokyo, Japan), which was administered intravenously to the patient in 25 s. An early arterial phase scan was initiated with bolus tracking of the abdominal aorta (AA) with a threshold of 40 HU and a delay of 8 s. Because a 70-keV VMI exhibits approximately the same contrast as an image with a tube voltage of 120 kV, which is commonly used in conventional CT, the 70-keV image was used as a standard [25].

Early arterial phase images were reconstructed in the axial plane by quantum iterative reconstruction (QIR; Siemens Healthineers) at the highest intensity level of 4 using the vascular reconstruction kernel (Bv44). The field of view was 345 mm and the matrix size was 512 × 512. The thickness was 0.4 mm without a gap during the early arterial phase. The 3D volume rendering (VR) images of the abdominal artery were reconstructed from the early arterial phase VMI (50 and 70-keV) using a 3D image analysis system (SYNAPSE VINCENT Version 6.7; Fujifilm, Tokyo, Japan).

Two radiologists (both with 11 years of abdominal imaging experience) independently placed regions of interest (ROIs) in the following areas: AA, celiac artery (CeA), superior mesenteric artery (SMA), RA, and right hepatic artery (RHA) and measured the CT attenuation at 50 and 70-keV VMI reconstructed from early arterial phase CT. The mean CT and standard deviation (SD) values for each artery measured by the two readers were averaged. Image noise (SD_fat) was defined as the SD of CT density of the anterior subcutaneous fat. The mean CT value of the muscles (HU_muscle) was defined as the CT value of the erector spinae muscles at the same slice and level as the CeA. The signal-to-noise ratio (SNR) and CNR were calculated as follows:

$$ \begin } = }\_}/}\_} \hfill \\ } = \left( }\_} - }\_}} \right)/}\_}. \hfill \\ \end $$

One radiologist (with 11 years of experience in abdominal imaging and specializing in 3D image processing) measured the length of the artery by analyzing the 3D VR images using the following steps:

Step 1 In the early arterial phase, VR images of the abdominal arteries were created at 50- and 70-keV VMI using SYNAPSE VINCENT. The observer was blinded to whether each CT image was a 50- or 70-keV CTA image.

Step 2 For the VR images, an automatic bone removal function was used to remove the bone and extract the entire artery of the upper abdomen.

Step 3 Cutting and vessel selection tools were used to manually extract the origin of the CeA to the periphery of the RHA, origin of the SMA to the periphery, and origin of the left RA to the periphery.

In cases in which the RHA bifurcated from the SMA (replaced RHA), we manually extracted VR images from the beginning of the SMA to the RHA. When extracting the renal artery, the opacity was adjusted to make the end of the renal artery visible, because the end of the renal artery was not visible at the default opacity because of the renal parenchyma. If an accessory RA was present, extraction was performed from the main RA to the periphery.

Step 4 In the vessels extracted in Step 3, the vessel pathways were automatically extracted using SYNAPSE VINCENT curved planar reformation (CPR) analysis.

Step 5 For the RHA measurements, the pathway from the CeA origin to the A8 periphery was selected from the vessel pathways, and the distance was measured automatically. For SMA measurements, the pathway was extracted using CPR analysis. The pathway from the origin of the SMA to the end of the ileocolic artery (ICA) was selected, and the distance was measured automatically. To measure the length of the RA, the path from the origin of the left RA to the periphery of the anterior superior segmental artery (ASSA) was selected and the distance was measured automatically.

3D images of the RHA, SMA, and RA length measurement methods are summarized in Figs. 2, 3, and 4, respectively. To determine which artery exhibited the greatest improvement in delineation at low-keV, we also calculated and compared the difference in the length of the RHA, SMA, and RA between 50 and 70-keV VMI.

3D images of a 60-year-old woman (a, b) and a 79-year-old man (c, d) for right hepatic artery (RHA) length measurement. (a, b) We performed curved planar reformation (CPR) analysis on cropped volume rendering (VR) images to extract vascular pathways on 50- and 70-keV virtual monoenergetic images (VMI). The purple line indicates that the software automatically extracted the vascular pathway and the yellow line indicates the pathway from the CeA origin to the A8 periphery. The results of the artery length measurements are shown in the bottom row. CeA, celiac artery; RAHA, right anterior hepatic artery; RPHA, right posterior hepatic artery. (c, d) We performed CPR analysis on the cropped VR images to extract the vascular pathways on 50- and 70-keV VMI. The results of the artery length measurements are displayed in the bottom row

3D images of a 75-year-old woman (a, b) and a 68-year-old man (c, d) for superior mesenteric artery (SMA) length measurement. (a, b) We performed curved planar reformation (CPR) analysis on the cropped volume rendering (VR) images to extract the vascular pathways on 50- and 70-keV virtual monoenergetic images (VMI). The purple line indicates that the software automatically extracted the vascular pathway, and the yellow line indicates the pathway from the SMA origin to the ileocolic artery (ICA) periphery. The results of the artery length measurements are displayed in the bottom row. MCA, middle colic artery; RCA, right colic artery. (c, d) CPR analysis was performed on the cropped VR images to extract the vascular pathways on 50- and 70-keV VMI. The results of the artery length measurements are displayed in the bottom row

3D images of a 61-year-old woman (a, b) and a 68-year-old man (c, d) for renal artery (RA) length measurement. (a, b) We performed curved planar reformation (CPR) analysis on the cropped volume rendering (VR) images to extract the vascular pathways on 50- and 70-keV virtual monoenergetic images (VMI). The purple line indicates that the software automatically extracted the vascular pathway, and the yellow line indicates the pathway from the RA origin to the anterior superior segment artery (ASSA) periphery. The results of the artery length measurements are displayed in the bottom row. (c, d) We performed CPR analysis on the cropped VR images to extract the vascular pathways on 50- and 70-keV VMI. The results of the artery length measurements are displayed in the bottom row.

Because all parameters were normally distributed using the Shapiro–Wilk test, parametric tests were performed. The SNR and CNR for each vessel at 50 and 70-keV VMI were compared using a paired t-test. The inter-reader reliability of the CT and SD values for each artery between the two observers was assessed by calculating intraclass correlation coefficients (ICCs) [26]. The lengths of each artery at 50 and 70-keV VMI were also compared using a paired t-test. The difference in arterial length at 50 and 70-keV VMI was compared between the three groups (CeA, SMA, and RA) using one-way analysis of variance (ANOVA) with post-hoc Tukey’s test. SPSS for Mac version 24 (IBM, Chicago, USA) was used for all statistical analyses. p values < 0.05 were considered to indicate a statistically significant difference.