Intrahepatic cholangiocarcinoma and hepatocellular carcinoma: differential diagnosis with contrast-enhanced ultrasound
- First Online:
- Cite this article as:
- Chen, L., Xu, H., Xie, X. et al. Eur Radiol (2010) 20: 743. doi:10.1007/s00330-009-1599-8
- 459 Views
We assessed the usefulness of contrast-enhanced ultrasound (CEUS) in the differentiation of intrahepatic cholangiocarcinoma (ICC) and hepatocellular carcinoma (HCC).
The CEUS enhancement patterns of 50 ICCs were retrospectively analysed and compared with 50 HCCs. Two readers independently reviewed the baseline ultrasound (BUS) and CEUS images and the diagnostic performances were evaluated by receiver operating characteristic (ROC) analysis. Time–intensity curves (TIC) were plotted for quantification analysis.
In the arterial phase, peripheral rim-like hyperenhancement, heterogeneous hyperenhancement, homogeneous hyperenhancement and heterogeneous hypoenhancement were found in 25, 10, 3 and 12 of the ICCs versus 2, 29, 19 and 0 of the HCCs (P < 0.001), respectively. The diagnostic performance of both readers in terms of the area under the ROC curve (0.745 vs. 0.933 for reader 1, and 0.803 vs. 0.911 for reader 2), sensitivity (28% vs. 90%, and 44% vs. 82%) and accuracy (64% vs. 90%, and 71% vs. 90%) improved significantly after CEUS (all P < 0.05). The interobserver agreement increased from κ = 0.575 at BUS to κ = 0.720 after CEUS. TICs demonstrated that the intensities of the peripheral and central portions of the ICCs were lower than those of HCCs (both P < 0.05).
CEUS improves the diagnostic performance significantly in the differentiation between ICC and HCC.
KeywordsContrast-enhanced ultrasoundIntrahepatic cholangiocarcinomaHepatocellular carcinomaDifferential diagnosisCharacterisation
Hepatocellular carcinoma (HCC) is the sixth most common neoplasm worldwide. Its incidence has increased in recent years and is currently the leading cause of death amongst cirrhotic patients [1, 2]. Intrahepatic cholangiocarcinoma (ICC) accounts for 5–30% of primary liver cancers, and its incidence rate has also increased in recent years . ICC and HCC have different biological behaviour and prognosis [1, 4–6]. Differentiation between them is essential for surgical planning and prognosis assessment. It is difficult for baseline ultrasound (BUS) to distinguish between them because ICC has no specific features [7–9]. Recent studies have proven that contrast-enhanced ultrasound (CEUS) improved the diagnostic performance in characterisation of focal liver lesions (FLLs) compared with BUS . However, so far there have been no studies evaluating the use of CEUS in differentiating between ICC and HCC. As shown in previous studies, ICC typically produces a peripheral rim-like enhancement in the arterial phase of CEUS [11, 12], whereas HCC is homogeneously or inhomogeneously hyperenhancing [13, 14]. We supposed that CEUS would therefore improve the ability to differentiate between these two entities. In this study, we compared the CEUS imaging characteristics of ICC and HCC, assessed the diagnostic value of CEUS in the differential diagnoses between them using receiver operating characteristic (ROC) analysis, and evaluated the difference in blood perfusion for these two entities quantitatively.
Materials and methods
Age range 18–80 years
Absence of severe cardiopulmonary diseases
Not pregnant or lactating
Lesions had not been treated before
Written informed consent was obtained from each patient and the study was approved by the institutional ethics committee. Final diagnoses included various types of lesions, such as HCC, ICC, haemangioma, metastatic liver cancer, focal nodular hyperplasia, and others.
Among the 3,407 patients, all 50 patients (34 men and 16 women; mean age 58 ± 11 years, age range 28–76 years) with pathologically proven ICCs were retrospectively enrolled as the study population for the ICC group. Five patients in the ICC group had a history of cirrhosis. Seven patients had multiple lesions (2 lesions in 5 patients, and more than 5 lesions in 2) and 43 patients had a solitary lesion in each. In patients with multiple lesions, only the largest lesion visualised on BUS was selected for CEUS evaluation. Thus, 50 ICCs were included in the study, and the maximum diameters of the ICCs were 2.0–15.5 cm (mean ± SD, 7.2 ± 2.8 cm).
Among the 3,407 consecutive patients who presented during the corresponding period, 875 patients had HCCs. Fifty sample cases were randomly selected to form the HCC group by using the sequence numbers created by the SPSS 13.0 software package (SPSS Inc, Chicago, IL). The diagnosis was confirmed either by pathological analysis (n = 38) or by clinical criteria (n = 12) according to the literature . The population comprised 43 men and 7 women (P < 0.05, compared with ICC) (mean age 53 ± 11 years, range 33–78 years. P < 0.02, compared with ICC). Twenty-six patients had a history of cirrhosis (P < 0.001, compared with ICC). Eleven patients had multiple lesions (2 lesions in 7 patients, 3 lesions in 1, 5 lesions in 1, and more than 5 lesions in 2) and 39 patients had a solitary lesion in each (P > 0.05, compared with ICC). In patients with multiple lesions, only the largest lesion was selected for CEUS evaluation. Therefore, 50 HCCs sized from 1.5 to 16.2 cm (mean ± SD, 6.1 ± 3.2 cm) (P > 0.05, compared with ICC) were observed.
BUS and CEUS examinations
Forty ICCs and 42 HCCs were examined with an Acuson Sequoia 512 (Siemens Medical Solutions, Mountain View, CA). The 4V1 vector transducer (frequency range 1.0–4.0 MHz) and low acoustic power real-time CEUS mode of contrast pulse sequencing (CPS; Siemens Medical Solutions) (mechanical index range 0.15–0.21) was applied.
Ten ICCs and 8 HCCs were examined with an Aplio XV (Toshiba Medical System, Tokyo, Japan), which was equipped with a 375BT convex transducer (frequency range 1.9–6.0 MHz) and contrast-specific mode of contrast harmonic imaging (CHI; Toshiba) with mechanical index of 0.05 to 0.08.
The contrast agent used was SonoVue® (BR1; Bracco SpA, Milan, Italy), a sulphur hexafluoride-filled microbubble that is covered by a phospholipid shell.
Two primary investigators, both with more than 8 years’ experience in liver BUS imaging and at least 5 years’ experience in liver CEUS, performed the imaging. They were not blinded to information on the patients’ clinical presentations and other imaging results. BUS was performed first. After identification of the target lesion the transducer was kept in a stable position while the imaging mode was changed to CEUS, and the mechanical index (MI) settings were adjusted to provide sufficient tissue cancellation with the maintenance of adequate depth penetration. Focus was positioned just below the bottom of the lesion and maintained the same position during the examination. The initial gain setting showed system noise in the far field of the contrast image that did not change during the whole examination. Then, a dose of 2.4 mL SonoVue was administered via the antecubital vein in a bolus fashion (within 1–2 s), followed by a flush of 5 mL of 0.9% normal saline by using a 20-gauge cannula. A stopwatch was started at the time of SonoVue administration. CEUS images were recorded continuously for a period of more than 240 s immediately after the injection, without any change in the machine settings. The arterial, portal and late phases were defined as 8–30 s, 31–120 s and 121–240 s after the injection, respectively. BUS images and CEUS cine clips were stored digitally on the hard disk in the imaging systems, and were transferred to a personal computer for subsequent analysis.
Immediately after the CEUS images were finished, all the BUS images and CEUS cine clips were analysed subjectively in consensus by the two primary investigators. The number, size and echogenicity of the lesion, as well as focal bile duct dilation around the tumour on BUS were recorded. On CEUS, the enhancement pattern in the arterial phase was classified as follows: peripheral rim-like enhancement i.e. a continuous ring of contrast enhancement seen at the periphery of the lesion; heterogeneous enhancement i.e. the whole lesion enhanced heterogeneously; homogeneous enhancement i.e. the entire lesion enhanced uniformly. The intratumoral blood vessels and tumour capsule that appeared during arterial phase were also evaluated. The enhancement level (hyper-, iso- or hypoenhancement, relative to the adjacent liver parenchyma) and their changes in the portal and the late phases, were evaluated.
Diagnostic criteria for ICC and HCC on baseline and contrast-enhanced ultrasound (US)
Baseline US criteria
Contrast-enhanced US criteria
Variable echogenicity, irregular lesion margin, peripheral bile duct dilatation
Peripheral irregular rim-like hyperenhancement, heterogeneous hypoenhancement or heterogeneous hyperenhancement during the arterial phase, hypoenhancement during the portal and late phases
Heterogeneous echogenic lesion, hypoechoic rim, peripheral or internal arterial flow signals
Homogeneous or heterogeneous hyperenhancement during the arterial phase, washout during the portal or late phase
One investigator who did not participate in the image analysis obtained still frames from the CEUS cine clips (every 5 s from 15 s to 60 s, and every 30 s from 60 s to 180 s) of all 50 ICC and 50 HCC records, and all images were transformed into the 8-bit tag image file format. The images were assessed by using the public domain image analysis program ImageJ (developed at the National Institutes of Health and available at http://rsb.info.nih.gov/nih-image/). Region of interest (ROI) was determined as a circular region of 1.0 cm (if lesion size ≥ 3.0 cm) or 0.5 cm (if lesion size < 3.0 cm) in diameter. They were placed in liver parenchyma and the peripheral and central regions of the tumour, respectively, in the same location on each image with the same depth, avoiding any large hepatic vessels and necrotic areas. Then, echo intensities of ROIs of each image were measured. A time–intensity curve (TIC) was plotted as mean intensity versus time.
The diagnostic performance of BUS and CEUS in differentiating ICC and HCC was assessed by an ROC curve, which was expressed as the area under the ROC curve (Az). The comparison of Az was evaluated by univariate z score test. If the lesion was assigned a confidence grade of 1 (definitely ICC) or 2 (probably ICC), the diagnosis was defined as a positive result; the lesion assigned a confidence grade of 3 (indeterminate), 4 (probably HCC) or 5 (definitely HCC) was defined as a negative result. The McNemar test or the Chi-squared test was used to evaluate the difference between qualitative data. P < 0.05 was considered to be statistically significant. The κ statistic was used to assess interobserver agreement. Agreement was graded as poor (κ < 0.20), moderate (κ = 0.20 to <0.40), fair (κ = 0.40 to <0.60), good (κ = 0.60 to <0.80), or very good (κ = 0.80–1.00). The statistical analyses were performed using the SPSS 13.0 software package (SPSS Inc, Chicago, IL).
On BUS, the numbers of ICC that showed hypo-, iso- and hyperechogenicity were 13 (26%), 8 (16%) and 29 (58%); for HCC, the corresponding numbers were 18 (36%), (16%) and 24 (48%) (P > 0.05). Intrahepatic bile duct dilation around the tumour was seen in 13 (26%) ICCs and 4 (8%) HCCs (P < 0.05).
In the portal phase, 36 (72%) ICCs that showed hyperenhancement in the arterial phase became hypoenhanced. Two (4%) ICCs that showed hyperenhancement in the arterial phase became isoenhanced. Twelve (24%) ICCs that showed hypoenhancement in the arterial phase remained hypoenhanced. Among 50 (100%) HCCs with hyperenhancement, 47 (94%) became hypoenhanced, 2 (4%) were isoenhanced and 1 (2%) remained hyperenhanced in the portal phase (P > 0.05, in comparison with ICC). In the late phase, all 50 ICCs and 50 HCCs were hypoenhanced.
Diagnostic performance of contrast-enhanced ultrasound for diagnoses of ICC in a patient group of 50 ICCs and 50 HCCs
Total (n = 100)
Comparisons of the correctly characterised lesions before and after contrast-enhanced ultrasound in 50 ICCs and 50 HCCs
Number of lesions correctly characterised
Reader 1 (%)
Reader 2 (%)
ICC (n = 50)
HCC (n = 50)
Intensity differences (dB) between different portions of the tumours (ICC and HCC) and surrounding liver parenchyma
In ICC, the central portion exhibited a low, plateau-like curve, and the peak intensity of 62 dB appeared at 25 s and 30 s. The central portion of HCC enhanced markedly from 71 dB at 15 s after contrast agent injection, and then decreased immediately after it reached the peak level (112 dB) (P < 0.05, compared with ICC) at 25 s (Fig. 6b). The mean intensity differences between ICC and adjacent liver parenchyma were lower than those between HCC and adjacent liver parenchyma at 25 s, 50 s, 90 s and 150 s after contrast agent injection (all P < 0.05).
It is difficult for BUS to differentiate ICC from HCC because the imaging findings of ICC on BUS are non-specific [7–9]. In recent years, CEUS has become an important technique for diagnosing FLLs and is increasingly accepted in clinical use . On CEUS, it is well known that a typical HCC is characterised by arterial phase hypervascularity and rapid washout in the late phase [16, 18, 19]. On the other hand, few studies with small numbers of cases reported the enhancement appearance of ICC on CEUS [11, 16, 20]. For example, von Herbay et al.  found that 1 ICC out of 126 lesions showed diffuse early arterial contrast enhancement followed by sharp demarcation of the lesion in the late phase. Quaia et al.  found that 1 ICC out of 452 liver lesions had a persistent heterogeneous and hypoechoic appearance. Xu et al.  reported three types of enhancement patterns in 18 ICCs: heterogeneous hyperenhancement, irregular peripheral rim-like hyperenhancement, and heterogeneous hypoenhancement. However, the differentiation between ICC and HCC by using CEUS has not been investigated.
In this study, four enhancement patterns of ICC in the arterial phase of CEUS were observed, which was consistent with the previous report . The most distinguishing feature was peripheral rim-like hyperenhancement, followed by heterogeneous hypoenhancement, heterogeneous hyperenhancement and homogeneous hyperenhancements. The enhancement patterns of ICC on CEUS may relate to the distribution of tumour vessels and pathological components in the tumour . The peripheral rim-like hyperenhancement may reflect more tumour vessels in the peripheral portion of ICC, as opposed to slender tumour vessels in the central portion . On the gross morphology of ICC, a large amount of fibrous stroma is frequently presented, particularly in the central area . When there are abundant tumour vessels and tumour cells diffused in the whole tumour, diffuse hyperenhancement may be shown on CEUS. In contrast, hypoenhancement may be present if the component of the whole tumour is marked fibrous stroma . In HCC, however, along with the stepwise progression of hepatocarcinogesis, an abnormal process of vascularisation by a shift from a venous to an arterial supply was demonstrated , hence 47 out of 50 HCCs in this study demonstrated the typical feature of homogeneous or heterogeneous hyperenhancement during the arterial phase and subsequent contrast washout in the portal phase.
The enhancement patterns of ICC and HCC correlated with tumour size. That is, most small ICCs and HCCs (≤3 cm) were homogeneously enhancing, while those larger than 3 cm enhanced heterogeneously or showed a peripherally enhancing rim. This reflected the intratumoral haemodynamic changes in the course of tumour growth. When an ICC is small, it may be abundant in tumour cells with few fibrous tissues, which leads to homogeneous hypervascularity similar to that of HCC . When grown large, more fibrous tissues and central necrosis appear, and the ICC would show peripheral hypervascularity or overall hypovascularity . Consequently, it is difficult to distinguish small ICC from HCC on CEUS, especially in those with homogeneous hyperenhancement.
During the portal phase, most ICCs and HCCs showed hypoenhancement, which indicated a predominant hepatic artery supply, but a decreased portal venous flow in both tumours. In the late phase, microbubbles washed out more conspicuously in all tumours compared with surrounding liver parenchyma [18, 29], and this feature has a high degree of sensitivity and specificity in the diagnosis of malignant FLLs [16, 30]. The detection rate of tumour vessels was lower in ICC (24%) in comparison with that in HCC (66%), which might be a clue in distinguishing between these two entities.
The analysis of CEUS enhancement patterns and their changes over time showed a possible benefit in differentiating ICC from HCC. The ROC analysis of our study confirmed that when compared with BUS, CEUS improved diagnostic performance, as shown by marked increases in Az values, sensitivity, accuracy and NPV. The number of indeterminate lesions reduced significantly after the contrast agent administration. Furthermore, it allowed for a better interobserver agreement, which revealed increased reproducibility of the diagnostic performance.
The low-MI CEUS allows the quantitative evaluation of blood flow in FLLs . In this study, the TIC analysis demonstrated marked arterial enhancement at the peripheral portion of both ICC and HCC after contrast medium injection, and clear washout during the portal and late phases. However, lower intensity and lower slope were confirmed in both the peripheral and central portions of ICC compared with those of HCC. In comparison with ICC, the abnormal artery supply of HCC was richer, and microvascularisation increased owing to tumour angiogenesis, as well as the presence of abnormal arterovenous shunts , so the intensity of HCC was higher than ICC in the arterial phase. On the other hand, the ICC washed out more thoroughly during the portal or late phase in comparison with HCC, both at the periphery and centre of the tumours, so that the intensity of ICC was lower than that of HCC. This phenomenon may indicate that more vessels that trapped microbubbles were present in the HCC.
The main limitation of the study was that some other frequently seen FLLs were not included in the study and some of them may be difficult to differentiate from ICCs and HCCs, such as metastatic liver carcinoma, abscess or other liver tumours [13, 33]. Therefore, the results of the study are only applicable when there is a need to make a distinction between ICC and HCC and the possibility of other tumours had already been excluded. This scenario is not rare in clinical practice after narrowing the range of differential diagnosis for FLLs. In this study, 5 ICCs which presented diffuse hyperenhancement in the arterial phase and washed out during the portal and late phases were misdiagnosed as HCC by both readers. And 2 HCCs with a non-enhancing area in the central portion and peripheral rim-like hyperenhancement in the arterial phase were incorrectly characterised as ICC. These findings may be related to the fact that both readers were blinded to the history of liver cirrhosis, elevation of serum biomarkers such as alpha-fetoprotein or CA 19-9, and other clinical presentations or laboratory findings. It was believed that access to this information would be helpful in solving these problems as it was shown that associated cirrhosis was more commonly found in the HCC group. Finally, as the number of small ICCs in our study is low, the usefulness of CEUS in diagnosing small ICCs needs further investigation with a large series of cases.
In summary, ICC has its own characteristics for the enhancement patterns in comparison with HCC on CEUS. CEUS improves the diagnostic performance significantly in the differentiation between ICC and HCC in comparison with BUS.
This work was supported in part by grant NCET-06-0723 from Chinese Ministry of Education and grant 2008-2-10 from the Public Welfare Research Special Project of the Chinese Ministry of Science and Technology.