Detection and differentiation of early hepatocellular carcinoma from cirrhosis using CT perfusion in a rat liver model

Jin-Ping Li, Guang-Long Feng, Da-Qing Li, Hai-Bo Wang, De-Li Zhao, Yong Wan and Hui-Jie Jiang

Harbin, China

Detection and differentiation of early hepatocellular carcinoma from cirrhosis using CT perfusion in a rat liver model

Jin-Ping Li, Guang-Long Feng, Da-Qing Li, Hai-Bo Wang, De-Li Zhao, Yong Wan and Hui-Jie Jiang

Harbin, China

BACKGROUND: Functional imaging such as CT perfusion can detect morphological and hemodynamic changes in hepatocellular carcinoma (HCC). Pre-carcinoma and early HCC nodules are difficult to differentiate by observing only their hemodynamics changes. The present study aimed to investigate hemodynamic parameters and evaluate their differential diagnostic cut-off between pre-carcinoma and early HCC nodules using CT perfusion and receiver operating characteristic (ROC) curves.

METHODS: Male Wistar rats were randomly divided into control (n=20) and experimental (n=70) groups. Diethylnitrosamine (DEN) was used to induce pre-carcinoma and early HCC nodules in the experimental group. Perfusion scanning was carried out on all survival rats discontinuously from 8 to 16 weeks. Hepatic portal perfusion (HPP), hepatic arterial fraction (HAF), hepatic arterial perfusion (HAP), hepatic blood volume (HBV), hepatic blood flow (HBF), mean transit time (MTT) and permeability of capillary vessel surface (PS) data were provided by mathematical deconvolution model. The perfusion parameters were compared among the three groups of rats (control, pre-carcinoma and early HCC groups) using the Kruskal-Wallis test and analyzed with ROC curves. Histological examination of the liver tissues with hematoxylin and eosin staining was performed after CT scan.

RESULTS: For HPP, HAF, HBV, HBF and MTT, there were significant differences among the three groups (P＜0.05). HAF had the highest areas under the ROC curves: 0.80 (control vs pre-carcinoma groups) and 0.95 (control vs early HCC groups) with corresponding optimal cut-offs of 0.37 and 0.42, respectively. The areas under the ROC curves for HPP was 0.79 (control vs pre-carcinoma groups) and 0.92 (control vs early HCC groups) with corresponding optimal cut-offs of 136.60 mL/min/100 mg and 108.47 mL/min/100 mg, respectively.

CONCLUSIONS: CT perfusion combined with ROC curve analysis is a new diagnosis model for distinguishing between pre-carcinoma and early HCC nodules. HAF and HPP are the ideal reference indices.

(Hepatobiliary Pancreat Dis Int 2016;15:612-618)

liver neoplasms;

computed tomography;

perfusion imaging;

rat model

Introduction

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer that poses a serious threat to human health.[1]Early detection and treatment of HCC are crucial to improving patients’ survival.[2]Therefore, it is important to have an accurate diagnosis of early HCC, so as to decide the accurate treatment. HCC occurs mainly in patients with cirrhosis, evolving from benign regenerative or dysplastic nodules. Dysplastic nodules are considered to be pre-carcinoma lesions.[3-6]Dysplastic nodules may be single or multiple, including low-grade and high-grade dysplastic nodules. High-grade dysplastic nodules have a high risk of malignant transformation. However, the detection of dysplastic nodules and its differentiation from early HCC by conventional CT examination are sometimes

very difficult, because of the complicated liver texture with cirrhotic background. It is also difficult to make a qualitative diagnosis through morphologic features alone even if it can be found.[7]

MRI has a very high resolution, but it is limited in the quantification of blood flow, because the signal enhancement of MRI does not show a linear correlation with concentration of the contrast medium.[8]Multi-slice CT (MSCT) perfusion has high temporal and spatial resolution, good reproducibility, simple protocol and the ability to measure hemodynamic parameters at the capillary level.[9-11]It is widely used in clinical and experimental research. At present, experimental rat models are often employed in studies of hemodynamic changes in liver cancer, because carcinogenesis in rats has a similar pathological process and characteristics to that in humans.[12,13]During the development of carcinogenesis, hepatic arterial blood supply increase gradually, which is different from normal liver parenchyma of portal vein blood supply.[14]The perfusion parameters can quantitatively analyze microcirculation hemodynamic changes and are helpful in the diagnosis of benign and malignant lesions in liver nodules. However, the critical diagnosis of pre-carcinoma and early HCC nodules is also hard with just perfusion. Receiver operating characteristic (ROC) curve analysis is a frequently used statistical analysis method, which has been widely employed by medical imaging researchers. It is mainly used for evaluating the diagnostic accuracy and providing diagnostic thresholds in medical research.

To the best of our knowledge, many studies[15-19]have reported changes in hepatic perfusion caused by cirrhosis and liver cancer in clinical and animal studies. However, the relationship between pre-carcinoma and early HCC nodules has not been studied. In the present study, we analyzed the hemodynamic characteristics of pre-carcinoma and early HCC nodules with CT perfusion and evaluated their diagnostic values for distinguishing precarcinoma from early HCC nodules using ROC curves. HCC group) and pre-carcinoma (pre-carcinoma group). The lesions were confirmed by histopathology. We closely monitored the rats daily. If the animals developed any sign of abnormality, such as reduced water intake, decreased activity, food consumption, and ruffled fur, the DEN injection was either delayed, terminated or the dose was reduced appropriately. From 8-16 weeks, the DEN injection dose changed to 25 mg/kg once a week. CT perfusion examination was performed twice a week from 8 to 16 weeks. If the definitive lesions were found in perfusion, the rats were killed, the perfusion was analyzed and liver tissues were examined histologically. This study was approved by the local ethics committee of Harbin Medical University for animal care and use. All the surgical procedures performed and the care given to the rats complied with institutional guidelines.

Methods

Animals and experimental design

Ninety male Wistar rats (two months old, body weight 180-200 g) were used in our study. The rats were given one week to adapt to the environment before the experiment, they were randomly divided into the control group (n=20) and experimental group (n=70). The rats in the control group were injected intraperitoneally (i.p) with saline; in the experimental group with 50 mg/kg of diethylnitrosamine (DEN, 0.95 g/mL Sigma, St. Louise, MO, USA) twice a week for 8 weeks to induce early HCC (early

MSCT imaging examination

All rats underwent liver CT examination, including plain CT and perfusion scans with MSCT. The clinical system CT was used in our study (Lightspeed 128-slice spiral CT; GE Healthcare, Milwaukee, WI, USA). After an overnight fast, the rats were anesthetized with an i.p injection of 3% sodium pentobarbital (2 mL/kg). The pinhead of a 1-mL syringe was used to establish venous channels through the tail vein. Then, the legs were fixed on cardboard with rubber band and pins in a supine position. Plain CT scans were first conducted with the following parameters: 5.0 mm slice thickness, 5.0 mm slice interval, 1.0 pitch, 120 kV tube voltage, 100 mA tube current, 512×512 matrix, and FOV 9.6 cm, volume coverage up to 40 mm. A slice with a clear image of the portal vein and the abdominal aorta was selected for CT perfusion. Diatrizoate with 57% concentration was given intravenously using a high-pressure syringe, 0.5 mL/sec injection rates and injection time of 6 seconds. The scanning time continued for 50 seconds. The other scan parameters were done using the same as protocol described above.

Image processing and data analysis

CT images were transferred to the GE AW4.5 workstation, and Perfusion 4 software was used for data processing. A deconvolution algorithm was used to fit the data of the time-density curve (TDC). Three regions of interest (ROIs) were carefully demarcated to obtain satisfactory TDC. One ROI (1.5 mm2) was in the aorta, the other ROI (1.5 mm2) was in the portal vein, while the third ROI (1.5 mm2) was in the lesion nodules or liver parenchyma in the control group. In each of the control subjects, ROIs were placed randomly in the liver parenchyma, but the part with large vessels was

avoided. Due to the influence of partial volume effects, the ROIs of the edge of the liver were also avoided. With this model, the following parameters were calculated in the ROIs: hepatic blood flow (HBF, mL/min/100 mg), hepatic blood volume (HBV, mL/100 mg), mean transit time (MTT, second), permeability of capillary vessel surface (PS, mL/min/100 mg), and hepatic arterial fraction (HAF). The HBF, hepatic arterial perfusion (HAP) and hepatic portal perfusion (HPP) were calculated using following equations: HBF=HAP+HPP; HAP=HBF×HAF; HPP=HBF×(1-HAF).

Histology

After imaging examination, the rats were killed by spinal dislocation, and liver samples were excised. The samples were fixed in 10% formaldehyde and embedded in paraffin for immunohistochemical staining. Consecutive 4 μm sections were cut and mounted on glass slides. The sections were consistent with CT images as much as possible. The sections were stained for histological evaluation using standard hematoxylin and eosin (HE) staining.

Statistical analysis

Data were expressed as mean±standard deviation (SD). All statistical analyses were carried out using the SPSS version 11.5 (SPSS Inc., Chicago, IL, USA). The perfusion parameters were compared among the three groups of rats (control group, pre-carcinoma group and early HCC group) using the Kruskal-Wallis test. The Mann-Whitney U test was used to determine the differences in perfusion parameters between the pre-carcinoma and early HCC nodules, and the control groups, respectively. The diagnostic performances of HBF, HBV, MTT, HAF and HPP were assessed with nonparametric ROC curves. ROC curves were used to determine the sensitivity, specificity, and corresponding cut-off value of each perfusion parameter.[20]A P value less than 0.05 was considered statistically significant.

Results

Establishment of the rat models of early HCC and pre-carcinoma

During the experimental period, rats in the control group were active and gained weight. Four weeks after DEN injection, there was no obvious difference between rats in the experimental group and those in the control group. However, most rats in the experimental group showed decreased activities, reduced food intake, significant weight loss and gray furs. Six rats died before the completion of the 8-week DEN injection; ten died between 8 and 16 weeks; nine died after anesthesia before CT scanning; five died during CT scan. Sixty rats were evaluated: 20 in the control group and 40 in the experimental group.

Histology

In the experimental group, a total of 44 pre-carcinoma nodules and 16 early HCC nodules were found in the 40 rats and the nodules diameter ranged from 5 to 15 mm. Cirrhotic livers with dysplastic nodules reached up to three-cell thick and exhibited reduced reticular structure of the nodule. Clustered steatosis, Mallory bodies and isolated adenoid structure of various sizes were also found in the nodules (Fig. 1A). Besides clustered steatosis, Mallory bodies and isolated adenoid structures, abnormal cancer cells or irregular hepatocyte plates, vascular invasion and false adenoid structure were also found in the early HCC nodules (Fig. 1B).

CT perfusion findings

CT perfusion was available and analyzed for the 60 rats. For all groups, the quality of the TDC was good (Fig. 2). All the perfusion parameters are summarized in Table 1 and Fig. 3. For HBF, HBV, MTT, HAF and

HPP, there were significant differences among the three groups (P＜0.05). However, no significant difference was found in the PS and HAP among the control group, precarcinoma and early HCC group (P＞0.05).

Fig. 1. The pathological images of pre-carcinoma and early HCC nodules in rat liver (HE staining, original magnification ×40). A: Image of dysplastic nodules with a higher magnification. The hepatocyte plates reached up to three-cell thick. B: Image of early HCC with a higher magnification. Images showed increased number of abnormal cells with higher nuclei-to-cytoplasm ratio.

Fig. 2. A: Early HCC original image of CT perfusion scan. White arrow indicated the nodule with high perfusion under the membrane of the liver. B: The corresponding TDC derived from analyses of ROIs (aorta and portal vein) for the production of perfusion parameters.

Compared with the control group, the pre-carcinoma nodules had significantly higher values of HAF (P＜0.05) and showed lower HBF, HBV and HPP (P＜0.05). There were no significant difference in the MTT, PS and HAP between the pre-carcinoma group and the control group (P＞0.05).

Compared with the control group, the early HCC group showed decreased HBF and HPP, and increased

MTT and HAF (P＜0.05), while no significant difference was found in HBV, PS and HAP (P＞0.05).

Table 1. Hemodynamic parameters in different ROIs in the control, pre-carcinoma and early HCC groups

Table 2. ROC curve analysis for HBF, HBV, MTT, HAF and HPP

Fig. 3. Box and Whisker plots of perfusion parameters in the control, pre-carcinoma, and early HCC groups.

The areas under the ROC curves and optimal cut-off of the HBF, HBV, MTT, HAF and HPP in differentiating between the control and pre-carcinoma groups, and between the control and early HCC groups are summarized in Table 2 and Fig. 4. HAF had maximum areas under the ROC curves (0.80 in control vs pre-carcinoma groups, 0.95 in control vs early HCC groups); corresponding optimal cut-off was 0.37 and 0.42, respectively. The areas under the ROC curves of HPP was 0.79 (control vs pre-carcinoma groups) and 0.92 (control vs early HCC groups), ranking second; corresponding optimal cut-off was 136.60 mL/min/100 mg and 108.47 mL/min/100 mg, respectively.

Fig. 4. ROC curves of HBF, HBV, MTT, HAF and HPP were used to differentiate the control group from the pre-carcinoma (A) and early HCC (B) groups.

Discussion

In our study, we used Wistar rats to study human liver hemodynamics of pre-carcinoma and early HCC nodules. It is generally accepted that Wistar rats can be used as ideal experimental models and this is supported by our pre-experiment and other reports.[14,21,22]

During DEN-induction of early HCC, we observed a significant weight loss caused by DEN toxicity in all the rats in the experimental groups. Sixteen out of the seventy rats died during the 16-week DEN injection, which proves that using this method to establish models has a high mortality rate. This view is consistent with previous report.[23]

Park et al[12]reported that in rats receiving an intraperitoneal injection of 50 mg/kg DEN twice a week, HCC developed from the promotion stage to the progression stage after 12 weeks of DEN treatment. In our study, we used the dosage of 50 mg/kg DEN twice a week for 8 weeks. The treatment dose and injection frequency were reduced appropriately after 8 weeks to control effectively the progress of early HCC development and thus form more pre-carcinoma and early HCC nodules. By doing this, we successfully achieved a total of 44 pre-carcinoma nodules and 16 early HCC nodules.

In the cirrhotic liver, various types of hepatocyte nodular lesions, ranging from regenerative nodules to HCC, are frequently found in the imaging examination. However, the qualitative diagnosis of these nodules is sometimes very challenging. Moreover, in the processing of carcinogenesis in the cirrhotic liver, before morphological changes occur, hemodynamic changes take place in the liver lesion. These changes cannot be discovered by routine examinations.[21]We tried to find an efficient method to resolve this issue. Currently, functional imaging, such as CT perfusion, which can also perform quantitative or semi-quantitative analysis of liver perfusion parameters in vivo, is being gradually used to reflect both the morphological and hemodynamic changes of liver tissues. The present study focused on evaluating the characteristics of the blood supply to pre-carcinoma and early HCC nodules.

The cirrhotic liver nodules transform into liver cancers through the transformation of capillary and angiogenesis in cirrhosis nodules.[3]These changes in pathological angiogenesis result in the changes in blood supply, which can be examined by CT perfusion. In normal livers, the hepatic artery accounts for 20%-25% and the portal vein accounts for 75%-80% of blood supply. The parenchyma of a normal liver is mainly provided by the portal vein, but in the processing of carcinogens, the blood supply is primarily through the hepatic artery. This means that the ratio of HAP to HPP is disrupted during the development of pre-carcinoma to early HCC nodules. This change is the basis for the judgment of deterioration. In this study, we observed increased HAP and decreased HPP during the development of early HCC, and these changes occurred gradually. But there was no significant difference in HAP among the control, pre-carcinoma and early HCC groups (P＞0.05), however, there was a

significant change in HPP, that is to say, portal vein hemodynamic changes was dominant. When there was a significant change in HAP, malignant transformation of nodules was likely to have completed. The decline in HPP is more than the increase in HAP, such that the rise of HAP compensates for the decrease in HPP. This is a self-regulating mechanism of the liver. The findings of our study are in line with the previous reports.[24-27]HAF reflects the change in the ratio of hepatic-arterial to portal-venous blood flow in liver lesions; there was an obvious gradual increase among the three groups in our study. HBF is representative of the total HAP and HPP; so, it will drop with a decrease in HPP. The PS indicated the level of vascular permeability and would be affected by the volume of nodules with the decrease of HBF, however, there was no significant change among the control group, pre-carcinoma and early HCC group and this result should be intensively studied.

It is generally accepted that nodules with increased arterial supply or decreased portal supply frequently undergo transformation to typical HCC over a short period.[28]Kudo et al[27]demonstrated that both dysplastic nodules and HCCs are mainly fed by portal flow and not arterial flow. It is difficult to differentiate between precarcinoma and early HCC nodules by their hemodynamics only. There is thus need to develop a new method for the differential diagnosis of pre-carcinoma and early HCC nodules.

ROC curve has been successfully used in radiology for describing the diagnostic accuracy of imaging methodologies and assessing new and existing techniques to see if diagnostic accuracy can be improved.[29,30]Evaluation of diagnostic performance is typically based on the ROC curve and the area under the ROC curve as its summary index.[31]In this study, we used it to evaluate the differential diagnosis value of each index for precarcinoma and early HCC nodules. Our results showed that HAF and HPP were the most relevant parameters, which had larger areas under the ROC curves, and their sensitivity and specificity were higher than those of other perfusion parameters. The findings of our study are in line with the report of Lee et al.[19]Moreover, we successfully obtained differential diagnosis cut-off between precarcinoma and early HCC nodules. To date, limited data are available regarding the use of both CT perfusion and ROC curves to determine differential diagnostic cut-off. It should be noted that there is a need to observe further the natural history of the pre-carcinoma to early HCC nodules, which may influence the accuracy of the pathological diagnosis and subsequent therapy.

Our study has some limitations. First, low-grade and high-grade dysplastic nodules are not completely separated. They have subtle different characteristics of circulation and histopathological changes. Second, the number of early HCC nodules was low. A study with a large number of early HCC nodules could improve the reliability of the results. Third, we did not compare the CT perfusion performance with another validated imaging technique like MRI. Finally, despite using the abdominal compression bandage, the CT perfusion examinations might have been influenced by the breathing motion of the rats, as this is an in vivo study.

In conclusion, the present study constructed a new model using CT perfusion and ROC curve statistics analysis for distinguishing pre-carcinoma nodules from early HCC nodules. HAF and HPP were ideal reference indices to distinguish them with the help of differential diagnostic cut-off. We believe that this diagnostic model is a valuable differential one, which will be useful in detecting early malignant nodules and therefore, has potential for clinical application.

Contributors: LJP and JHJ proposed the study and wrote the first draft. LJP, FGL and LDQ performed the research. WHB, ZDL and WY collected and analyzed the data. All authors contributed to the design and interpretation of the study and to further drafts. JHJ is the guarantor.

Funding: This study was supported by grants from the National Natural Science Foundation of China (81301275, 81471736 and 81671760), the National Science and Technology Pillar Program during the Twelfth Five-Year Plan Period (2015BAI01B09), and Heilongjiang Province Foundation for Returness (LC2013C38).

Ethical approval: This study was approved by the local ethics committee of Harbin Medical University for animal care and use. Competing interest: No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

1 Kim BK, Kim SU, Park JY, Kim DY, Ahn SH, Park MS, et al. Applicability of BCLC stage for prognostic stratification in comparison with other staging systems: single centre experience from long-term clinical outcomes of 1717 treatment-naïve patients with hepatocellular carcinoma. Liver Int 2012;32:1120-1127.

2 Yang L, Parkin DM, Ferlay J, Li L, Chen Y. Estimates of cancer incidence in China for 2000 and projections for 2005. Cancer Epidemiol Biomarkers Prev 2005;14:243-250.

3 Efremidis SC, Hytiroglou P. The multistep process of hepatocarcinogenesis in cirrhosis with imaging correlation. Eur Radiol 2002;12:753-764.

4 Iavarone M, Manini MA, Sangiovanni A, Fraquelli M, Forzenigo LV, Di Tommaso L, et al. Contrast-enhanced computed tomography and ultrasound-guided liver biopsy to diagnose dysplastic liver nodules in cirrhosis. Dig Liver Dis 2013;45:43-49.

5 Park YN, Kim MJ. Hepatocarcinogenesis: imaging-pathologic correlation. Abdom Imaging 2011;36:232-243.

6 Hussain SM, Reinhold C, Mitchell DG. Cirrhosis and lesion characterization at MR imaging. Radiographics 2009;29:1637-1652.

7 Jin GZ, Dong H, Yu WL, Li Y, Lu XY, Yu H, et al. A novel panel of biomarkers in distinction of small well-differentiated HCC from dysplastic nodules and outcome values. BMC Cancer 2013;13:161.

8 Brasch RC, Weinmann HJ, Wesbey GE. Contrast-enhanced NMR imaging: animal studies using gadolinium-DTPA complex. AJR Am J Roentgenol 1984;142:625-630.

9 Petralia G, Fazio N, Bonello L, D’Andrea G, Radice D, Bellomi M. Perfusion computed tomography in patients with hepatocellular carcinoma treated with thalidomide: initial experience. J Comput Assist Tomogr 2011;35:195-201.

10 Ippolito D, Sironi S, Pozzi M, Antolini L, Invernizzi F, Ratti L, et al. Perfusion CT in cirrhotic patients with early stage hepatocellular carcinoma: assessment of tumor-related vascularization. Eur J Radiol 2010;73:148-152.

11 Jiang HJ, Zhang ZR, Shen BZ, Wan Y, Guo H, Li JP. Quantification of angiogenesis by CT perfusion imaging in liver tumor of rabbit. Hepatobiliary Pancreat Dis Int 2009;8:168-173.

12 Park DH, Shin JW, Park SK, Seo JN, Li L, Jang JJ, et al. Diethylnitrosamine (DEN) induces irreversible hepatocellular carcinogenesis through overexpression of G1/S-phase regulatory proteins in rat. Toxicol Lett 2009;191:321-326.

13 Schiffer E, Housset C, Cacheux W, Wendum D, Desbois-Mouthon C, Rey C, et al. Gefitinib, an EGFR inhibitor, prevents hepatocellular carcinoma development in the rat liver with cirrhosis. Hepatology 2005;41:307-314.

14 Ma G, Bai R, Jiang H, Hao X, Ling Z, Li K. Assessment of hemodynamics in a rat model of liver cirrhosis with precancerous lesions using multislice spiral CT perfusion imaging. Biomed Res Int 2013;2013:813174.

15 Bayraktutan U, Kantarci A, Oğul H, Kizrak Y, Özyiğit Ö, Yüceler Z, et al. Evaluation of hepatocellular carcinoma with computed tomography perfusion imaging. Turk J Med Sci 2014;44:193-196.

16 Wu D, Tan M, Zhou M, Sun H, Ji Y, Chen L, et al. Liver computed tomographic perfusion in the assessment of microvascular invasion in patients with small hepatocellular carcinoma. Invest Radiol 2015;50:188-194.

17 Fischer MA, Kartalis N, Grigoriadis A, Loizou L, Stål P, Leidner B, et al. Perfusion computed tomography for detection of hepatocellular carcinoma in patients with liver cirrhosis. Eur Radiol 2015;25:3123-3132.

18 Wang L, Fan J, Ding X, Sun J, Zhang M. Assessment of liver fibrosis in the early stages with perfusion CT. Int J Clin Exp Med 2015;8:15276-15282.

19 Lee DH, Lee JM, Klotz E, Han JK. Multiphasic dynamic computed tomography evaluation of liver tissue perfusion characteristics using the dual maximum slope model in patients with cirrhosis and hepatocellular carcinoma: a feasibility study. Invest Radiol 2016;51:430-434.

20 Pepe MS, Feng Z, Janes H, Bossuyt PM, Potter JD. Pivotal evaluation of the accuracy of a biomarker used for classification or prediction: standards for study design. J Natl Cancer Inst 2008;100:1432-1438.

21 Bai RJ, Wang JE, Jiang HJ, Hao XJ, Dong XP, Huang YH, et al. Investigation on the optical scan condition for imaging of multi-slice spiral CT liver perfusion in rats. Chin Med J (Engl) 2013;126:4742-4746.

22 Martins PN, Neuhaus P. Surgical anatomy of the liver, hepatic vasculature and bile ducts in the rat. Liver Int 2007;27:384-392.

23 Schneider C, Teufel A, Yevsa T, Staib F, Hohmeyer A, Walenda G, et al. Adaptive immunity suppresses formation and progression of diethylnitrosamine-induced liver cancer. Gut 2012;61: 1733-1743.

24 Tsushima Y, Blomley JK, Kusano S, Endo K. The portal component of hepatic perfusion measured by dynamic CT: an indicator of hepatic parenchymal damage. Dig Dis Sci 1999;44:1632-1638.

25 Solt DB, Hay JB, Farber E. Comparison of the blood supply to diethylnitrosamine-induced hyperplastic nodules and hepatomas and to the surrounding liver. Cancer Res 1977;37:1686-1691.

26 Conway JG, Popp JA, Thurman RG. Microcirculation of hepatic nodules from diethylnitrosamine-treated rats. Cancer Res 1985;45:3620-3625.

27 Kudo M, Hatanaka K, Inoue T, Maekawa K. Depiction of portal supply in early hepatocellular carcinoma and dysplastic nodule: value of pure arterial ultrasound imaging in hepatocellular carcinoma. Oncology 2010;78:60-67.

28 Matsui O, Kobayashi S, Sanada J, Kouda W, Ryu Y, Kozaka K, et al. Hepatocelluar nodules in liver cirrhosis: hemodynamic evaluation (angiography-assisted CT) with special reference to multi-step hepatocarcinogenesis. Abdom Imaging 2011;36:264-272.

29 Thompson JD, Manning DJ, Hogg P. The value of observer performance studies in dose optimization: a focus on free-response receiver operating characteristic methods. J Nucl Med Technol 2013;41:57-64.

30 Gönen M. Mixtures of receiver operating characteristic curves. Acad Radiol 2013;20:831-837.

31 Ma H, Bandos AI, Gur D. On the use of partial area under the ROC curve for comparison of two diagnostic tests. Biom J 2015;57:304-320.

Received May 15, 2016

Accepted after revision August 31, 2016

Author Affiliations: Department of Radiology, Second Affiliated Hospital, Harbin Medical University, Harbin 150086, China (Li JP, Feng GL, Li DQ, Wang HB, Zhao DL, Wan Y and Jiang HJ)

Hui-Jie Jiang, MD, PhD, Department of Radiology, Second Affiliated Hospital, Harbin Medical University, Harbin 150086, China (Tel: +86-451-86605576; Email: jhjemail@163.com)

© 2016, Hepatobiliary Pancreat Dis Int. All rights reserved.

10.1016/S1499-3872(16)60148-0

Published online November 4, 2016.