This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Ethics Committee of Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine (Approval No, XHEC-F-2013-016). All surgery was performed under ketamine hydrochloride and xylazine anesthesia, and animal suffering was minimized.

Twelve adult New Zealand white rabbits (Xinhua Hospital Laboratory Animal Center) weighing between 2.88 and 3.15 kg were implanted with a hepatic VX2 tumor using methodology similar to Kim et al[10]. Three weeks after implantation, the tumor size was expected to be approximately 3 cm. Non-contrast enhanced CT scanning was performed for the 12 experimental rabbits to evaluate the tumor size. No tumor was observed in the liver of one experimental rabbit. The tumor sizes for the remaining eleven rabbits were 3.3 ± 0.43 cm, ranging from 2.8 to 3.8 cm. One daughter nodule (0.51 cm in size) adjacent to the main tumor was detected in the liver of one rabbit. The eleven main tumors were evaluated in the perfusion CT study.

Before perfusion CT scan, each rabbit was anesthetized by intramuscular injection of 25 mg/kg body weight ketamine hydrochloride (Ketamine, Fujian, China) and 0.1 mL/kg 2% xylazine (Lu mian ning, Jilin, China). A 24-gauge cannula was inserted into the auricular vein. The rabbit was fixed on a board in supine position and properly covered with a bandage on the belly to minimize the respiratory motion.

Eleven rabbits were examined with a 256-slice CT scanner (Brilliance iCT; Philips Healthcare, Cleveland, Ohio). Repeated perfusion CT scanning of the upper abdomen was performed by using two different protocols: (1) conventional tube potential (120 kVp, 100 mAs) and concentration contrast medium (5 mL omnipaque at 350 mgI/mL); and (2) low tube potential (80 kVp, 80 mAs) and low concentration contrast medium (4 mL, visipaque at 270 mgI/mL). We used 80 kVp and 80 mAs setting as low radiation dose scanning protocol, based on Verdun et al[11] recommendation regarding low radiation dose CT scan protocol for pediatric patients with body weight from 2.5 to 5 kg. The dedicated perfusion scanning protocol with 8 cm wide coverage in one gantry rotation was used for whole liver perfusion CT. A 24-h interval was allowed between the use of low and conventional dose protocols for each rabbit. Iohexol (Omnipaque 350 mgI/mL, GE healthcare, Shanghai, China) and Iodixanol (visipaque, 270 mgI/mL, GE healthcare, Cork, Ireland) for conventional and low dose protocols, respectively, were administered through the auricular vein with a power injector (Mallinckrodt, Liebel-Flarseim Company, United States) at the rate of 1.5 mL/s. The perfusion CT scanning of the entire liver began at 2 s before bolus injection of contrast medium and continued for 90 s (gantry cycle time: 1.5 s, gantry cycle: 60 times). The scanning parameters were as follows: hepatic perfusion mode with field of view (FOV) 15 cm; slice thickness, 1.5 mm; no gap; matrix size, 512 × 512.

Conventional 120 kVp (protocol A) image data were reconstructed using the filtered back projection (FBP) algorithm. Images obtained from 80 kVp imaging were reconstructed by using the hybrid iterative reconstruction algorithm (protocol B) and FBP (protocol C). The level of hybrid iterative reconstruction was selected as 40% that theoretically provides 40% noise reduction for abdominal CT image data.

The perfusion images were transferred to the dedicated post-processing workstation (Extended Brilliance Workspace, V4.5.3, Philips Healthcare) for functional analysis using the maximum slope method. The region of interest (ROI) was placed on the abdominal aorta at the level of hepatic hilum, the main portal vein and the cortex of right kidney, to generate time-density curve (TDC). The rabbit’s relative small spleen cannot be easily identified; therefore, we selected the right renal cortex to differentiate the arterial and portal venous phases[12]. A radiologist (YFC) with 4 years of abdominal CT interpretation experience manually drew the ROI, outlining the contour of the whole tumor as well as the portion of viable tumor with the most marked enhancement. All measurements were performed three times. Three ROIs each with 50 mm² area were drawn on the hepatic lobe without tumor, avoiding the large vessels. Each ROI on the 120 kVp images was matched with that on 80 kVp images as much as possible. Subsequently, hepatic blood flow (HBF), hepatic arterial perfusion (HAP), hepatic portal perfusion (HPP), and hepatic perfusion index (HPI) for the whole tumor, viable tumor and normal liver parenchyma were automatically calculated with protocols A, B and C.

Three ROIs each with 50 mm² area were drawn on the hepatic parenchyma without tumor, avoiding large vessels at the level of maximum tumor diameter. The normal liver attenuation (ROI_liver) was defined as the mean CT value of the three ROIs. The mean tumor attenuation (ROI_tumor) was assessed by manually placing circle or ovoid ROIs as much as possible to encompass the enhanced portion of the tumor in three consecutive scanning slices. The lesion-to-liver contrast-to-noise ratio (CNR) was calculated using the following equation: CNR = (ROI_tumor-ROI_liver)/SD_noise, where ROI_liver is the mean attenuation of normal hepatic parenchyma; ROI_tumor is the mean attenuation of viable tumor component; and SD_noise is measured as the standard deviation of the pixel values from a circular or ovoid ROI drawn in the subcutaneous fat of the abdominal wall at the level of maximum tumor diameter[13]. All measurements were performed in both the hepatic arterial and portal venous phases; namely, CNR in the arterial phase (CNR_HAP) and portal venous phase (CNR_PVP) were both obtained. The timings of arterial and portal venous phases were defined as 11 s and 25 s, respectively, after enhancement in the aorta reached 100 HU[10].

After perfusion CT with low and conventional dose protocols, rabbits were sacrificed by injection of xyazine overdose (Lu mian ning, Jilin, China). Then, liver tissues were sliced into 1 mm thick sections in the orientation matching the CT scanning. The specimens were fixed with formalin for 24 h, embedded in paraffin and stained with hematoxylin and eosin, anti-CD31 and anti-vascular endothelial growth factor (VEGF) antibodies. The histological characterization of tumor was assessed by a pathologist (XY W).

All numeric values are expressed as mean ± standard deviation (SD). To compare CT perfusion parameters in whole tumor, viable tumor, and normal liver parenchyma for protocols A, B and C, one-way analysis of variance (ANOVA) was used. P < 0.05 was considered statistically significant. If a statistically significant difference was obtained, the Dunnett test was performed with protocol A serving as a control. Correlations among perfusion parameters for the whole tumor, viable tumor, and normal liver parenchyma measured in protocols A and B were assessed by Pearson correlation analysis. A correlation coefficient (r) between 0.5 and 0.8 indicated a moderate positive relationship; r ranging from 0.81 to 1.0 indicated a strong positive relationship[14]. Bland-Altman plots were employed to evaluate the agreement of perfusion parameters between protocols A and B. Dunnett test was used to compare the noise, CNR, and CT values for ROI in the three sets of perfusion CT, with protocol A used as a control. All statistical analyses were performed with SPSS version 15.0 for Microsoft Windows (SPSS, Inc., Chicago, IL) and the Medcalc software (Ostend, Belgium).

Perfusion CT was successfully performed in all 11 rabbits (Figure 1). Figure 2 shows that perfusion parameters for both viable tumor and whole tumor were significantly higher compared with those obtained for the normal liver parenchyma with protocols A and B, respectively (P < 0.05). However, no significant differences in HPP among viable tumor, whole tumor, and normal liver parenchyma were obtained using protocols A and B (P = 0.421 and 0.251, respectively). Therefore, a Dunnett test was not performed for this group.

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Computed tomography perfusions using both 80 kVp protocol with iterative reconstruction (4 mL Visipaque 270, 354. 1 mgI/kg) and 120 kVp protocol with filtered back projection (5 mL Omnipaque 350, 573.8 mgI/kg) at a 24-h interval were performed for liver VX2 tumor in a rabbit (body weight, 3.05 kg). A-D: Pseudocolor perfusion maps with the 80 kVp protocol and iterative reconstruction for HAP, HPP, HBF and HPI, respectively; E-H: Pseudocolor perfusion maps with the 120 kVp protocol and FBP for HAP, HPP, HBF, and HPI, respectively; I,J: Arterial phase images obtained using the 80 kVp protocol with iterative reconstruction and 120 kVp protocol reveal a markedly enhanced mass with central necrotic change, respectively. Computed tomography perfusion maps obtained using low tube potential with low concentration contrast medium depict a hypervascular mass with central hypovascular component, which is comparable to that obtained using the conventional tube potential and concentration contrast medium protocol. IR: Iterative reconstruction; FBP: Filtered back projection; HAP: Hepatic arterial perfusion; HPP: Hepatic portal perfusion; HBP: Hepatic blood flow; HPI: Hepatic perfusion index.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Box-and-whisker plots of hepatic arterial perfusion (A), hepatic portal perfusion (B), hepatic blood flow (C) and hepatic perfusion index (D) values were obtained with the 80 kVp protocol and iterative reconstruction and 120 kVp protocol to assess the difference in perfusion parameters among normal liver parenchyma, viable tumor and whole tumor (Donnut test, normal liver parenchyma set as control). Center lines in box represent mean HAP, HPP, HBF and HPI values with upper and lower box margins denoting maximum and minimum percentile of values, respectively. Whiskers gives the range of values. IR: Iterative reconstruction; HAP: Hepatic arterial perfusion; HPP: Hepatic portal perfusion; HBP: Hepatic blood flow; HPI: Hepatic perfusion index; HBF: Hepatic blood flow.

Similarly, analysis of variance showed that there were no significant differences among protocols A, B and C with regard to HAP, HPP, HBF and HPI for normal liver, viable tumor component and whole tumor (P > 0.05; Table 1).

HAP (r = 0.577, P = 0.031), HPP (r = 0.527, P = 0.048), HBF (r = 0.583, P = 0.03), and HPI (r = 0.531, P = 0.046) for normal liver parenchyma at 80 kVp with the iterative reconstruction protocol showed a moderate positive correlation with values measured at 120 kVp.

As shown in Figure 3, strong positive correlations were obtained in HAP (r = 0.907; P < 0.001), HPP (r = 0.819; P = 0.002) and HBF (r = 0.945; P < 0.001) values for viable tumor between protocols A and B. HPI (r = 0.736; P = 0.01) for viable tumor determined with protocol A showed a moderate positive correlation with the value measured with protocol B. Bland-Altman plots for viable tumor yielded mean differences of -3.7 for HAP (95% limits of agreement, -45.5 to 38 mL/min per 100 mg), 1.2 for HPP (95% limits of agreement, -28.4 to 30.8 mL/min per 100 mg), -2.6 for HBF (95% limits of agreement, -50.0 to 44.8 mL/min per 100 mg), and 1.0 for HPI (95% limits of agreement, -15.2% to 13.3%), which did not differ significantly from zero. In clinical interpretation, the 95% limits of agreement may lead to misinterpretation of tumor angiogenesis and liver perfusion in few cases, and are therefore considered to be a moderately acceptable agreement. For example, all points were located within 95% limits of agreement values (Figure 3), and HPI at 80 kVp with iterative reconstruction differed from HPI at 120 kVp by 15.2% below and 13.3% above; with mean value of these two HPIs being 56.16%, the misinterpretation of tumor perfusion may occur in few atypical cases.

Open in New Tab   Full Size Figure   Download Figure

Figure 3 Correlation and agreement of perfusion parameters for viable tumor component obtained using the 80 kVp protocol with iterative reconstruction (low radiation dose and concentration contrast medium) and 120 kVp protocol with filtered back projection (conventional radiation dose and concentration contrast medium). A: HAP; B: HPP; C: HBF; D: HPI; E: HAP-HAP; F: HPP-HPP; G: HBF-HBF; H: HPI-HPI. IR: Iterative reconstruction; HAP: Hepatic arterial perfusion; HPP: Hepatic portal perfusion; HBF: Hepatic blood flow; HPI: Hepatic perfusion index.

Similarly, correlation and agreement data among perfusion parameters for whole tumor between protocols A and B are shown in Figure 4. HAP, HPP and HBF determined with protocol A showed a strong positive correlation with values measured at 80 kVp. There was a moderate positive correlation for HPI of whole tumor between protocols A and B. Bland-Altman plots suggested a moderate acceptable agreement for these perfusion parameters between protocols A and B.

Open in New Tab   Full Size Figure   Download Figure

Figure 4 Correlation and agreement of perfusion parameters for whole tumor obtained using the 80 kVp protocol with iterative reconstruction (low radiation dose and concentration contrast medium) and 120 kVp protocol with filtered back projection (conventional radiation dose and concentration contrast medium). A: HAP; B: HPP; C: HBF; D: HPI; E: HAP-HAP; F: HPP-HPP; G: HBF-HBF; H: HPI-HPI. IR: Iterative reconstruction; HAP: Hepatic arterial perfusion; HPP: Hepatic portal perfusion; HBF: Hepatic blood flow; HPI: Hepatic perfusion index.

As shown in Table 2, the CT values obtained for normal liver parenchyma in arterial and portal venous phases with protocols B and C were significantly different from those obtained with protocol A (P < 0.05). However, there was no significant difference among protocols A, B and C for the same parameters of viable tumor in arterial and portal venous phases, respectively (P > 0.05). Image noises in arterial and portal venous phases with protocol C were 70% and 63% higher than values obtained with protocol A (13.85 ± 1.05 HU vs 8.16 ± 0.95 HU, P < 0.001; 13.47 ± 1.27 HU vs 8.27 ± 1.17 HU, P < 0.001, respectively). In comparison with image noises determined by protocol C, 80 kVp with iterative reconstruction reduced 33% and 29% of image noises in arterial and portal venous phases (9.31 ± 0.94 HU vs 13.85 ± 1.05 HU; 9.50 ± 0.93 HU vs 13.47 ± 1.27 HU, respectively). However, there was a significant difference in image noises during arterial and portal venous phases between protocols B and A (9.31 ± 0.94 HU vs 8.16 ± 0.95 HU, P = 0.019; 9.50 ± 0.93 HU vs 8.27 ± 1.17 HU, P = 0.031). There was no significant difference between protocols B and A in tumor-to-liver CNR values during both arterial and portal venous phases (P = 0.814, P = 0.587), while CNR values obtained with protocol C significantly differed from those measured with protocol A (P = 0.048, P = 0.004).

The mean values of contrast medium dose delivered for the 80 kVp and 120kVp protocols were 360.1 ± 13.3 and 583.5 ± 21.5 mgI/kg, respectively (P < 0.001). Low tube potential allowed a 38% decrease in contrast medium dose in comparison with conventional tube potential (Table 3).

Figure 5 illustrates the histopathological features of hepatic VX2 tumor in rabbits. The epithelial cells in tumor microvessels were stained brown with anti-CD31 and protoplasm of tumor cells was stained brown with anti-VEGF. In addition, microvessels were heterogeneously distributed within the tumor and showed a predominantly peripheral distribution. These findings were correlated with perfusion CT images.

Open in New Tab   Full Size Figure   Download Figure

Figure 5 Histological and immunohistochemical micrographs of hepatic VX2 tumor. A: Tumor tissue stained with hematoxylin and eosin (magnification × 400) demonstrates that the normal hepatic sinusoidal capillary disappeared and was replaced by a large amount of tumor cells; B: CD31 staining of tumor tissue shows the formation of new microvessels with incomplete layer of endothelial wall (arrows, magnification × 400); C: Vascular endothelial growth factor staining of tumor tissue depicts positive expression in the cytoplasm of tumor cell (arrows, magnification × 400).

Radiation doses were automatically measured by the CT scanner during scanning. The radiation dose for the 80 kVp group was 73% lower than that used in the 120 kVp group (CTDIvol: 237.31 mGy, DLP: 1898.5 mGy • cm vs CTDIvol 868.98 mGy, DLP 6951.8 mGy • cm).

Computed tomography (CT) perfusion plays an important role in quantitatively assessing the perfusion of abdominal organs and lesions. However, concerns regarding the relatively high radiation dose and contrast-induced nephropathy limited its wide use. There were rarely reported data on simultaneously implementing low radiation dose and low contrast medium concentration protocol for hepatic CT perfusion.

The technique of low tube potential combined with iterative reconstruction has emerged as a robust tool to dramatically decrease the radiation dose without compromising image quality and increase the conspicuity of hypervascular tumor during contrast enhanced CT scans. Therefore, the author aimed to evaluate whether hepatic CT perfusion with low tube potential and low contrast medium concentration protocol could provide similar perfusion parameters of rabbit VX2 tumor in comparison with conventional protocol.

This study reported for the first time that CT perfusion with low dose protocol achieved a 38% reduction in contrast medium consumption and a 73% reduction in radiation dose without significant influence on the hepatic perfusion parameters in rabbit VX2 tumor.

It is promising that hepatic CT perfusion with the combination of low tube potential with iterative reconstruction and low concentration contrast medium may be clinically used to decrease radiation dose and the risk of contrast-induced nephropathy without significant influence on perfusion parameters.

Since low tube potential is closer to the iodine K-edge of 33 keV, this technique increases the contrast of iodine and may allow for a reduction in the administered contrast medium dose.

Authors evaluate the feasibility of low radiation dose and low concentration contrast medium with iterative reconstruction on a 256-slice CT for hepatic VX2 tumor in rabbits, and conclude that perfusion CT with low tube potential and low concentration contrast agent can dramatically decrease radiation dose and image noise with similar conspicuity of tumor compared to conventional tube potential with conventional concentration contrast medium and does not significantly influence perfusion parameters for liver VX2 tumor in rabbits. This is an important study and if it could be repeated in human subjects would improve the safety of CT scanning for patients and allow the technique to be used in patients who would otherwise be contraindicated because of potential toxicity.