- Email: [email protected]
Is gadoxetic acid-enhanced MRI limited in tumor characterization for patients with chronic liver disease?
Magnetic Resonance Imaging 32 (2014) 1214–1222
Contents lists available at ScienceDirect
Magnetic Resonance Imaging journal homepage: www.mrijournal.com
Is gadoxetic acid-enhanced MRI limited in tumor characterization for patients with chronic liver disease?☆ Soyi Kwon a, Young Kon Kim a,⁎, Hyun Jeong Park b, Woo Kyoung Jeong a, Won Jae Lee a, Dongil Choi a a b
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea Department of Radiology, Chung-Ang University Hospital, Seoul, Republic of Korea
a r t i c l e
i n f o
Article history: Received 7 April 2014 Revised 9 June 2014 Accepted 10 August 2014 Keywords: Hepatocellular carcinoma Gd-EOB-DTPA Diffusion-weighted imaging Tumor characterization Cholangiocarcinoma
a b s t r a c t Purpose: There are pros and cons to the use of gadoxetic acid in hepatocellular carcinoma (HCC) workup due to the potential for high false positive diagnosis. This study was conducted to investigate the preoperative diagnostic performance of gadoxetic acid-enhanced MRI protocol including diffusionweighted imaging (DWI) with emphasis on tumor characterization developed in high risk HCC patients. Materials and methods: We included 144 patients (102 men, 42 women; age range 33–74 years) with chronic viral hepatitis or cirrhosis and 183 focal hepatic tumors (size range, 0.4–11.0 cm; mean, 3.2 cm), including 148 HCCs, 13 cholangiocarcinomas, 12 hemangiomas, three hepatocellular adenomas, two focal nodular hyperplasias, and five other tumors. All patients underwent gadoxetic acid-enhanced MRI protocol with DWI. MRIs were independently interpreted by three observers for the detection and characterization of hepatic tumors. Results: Sensitivities for detecting all 183 liver tumors were 98.4%, 97.8%, and 96.2% for each observer, respectively, with a 97.5% for pooled data. Among 183 hepatic tumors, 91.3% (n = 167), 87.4% (n = 160), and 86.9% (n = 159) were correctly characterized according to their reference standard by each observer, respectively. In 13 cholangiocarcinomas, one to three were misinterpreted as HCC, and the remaining tumors were correctly characterized by each observer. The accuracies (Az) of MRI for HCC diagnosis were 0.952 for observer 1, 0.906 for observer 2, and 0.910 for observer 3, with 0.922 for pooled data. There was good inter-observer agreement. Conclusion: The gadoxetic acid-enhanced MRI including DWI showed a reasonable performance for tumor characterization with high sensitivity for tumor detection in patients with chronic liver disease, despite concerns of high false positive diagnosis of hypervascular tumors. © 2014 Elsevier Inc. All rights reserved.
1. Introduction The goal of liver imaging for patients with chronic liver disease is the early detection and accurate characterization of hepatocellular carcinoma (HCC) by differentiating it from benign cirrhosisassociated hepatocellular nodules as well as other hepatic tumors. This increases the success rate of curative treatment and leads to a better outcome [1,2]. Magnetic resonance imaging (MRI) has the potential to better fulfill these imaging requirements than other liver imaging modalities since it not only offers excellent soft tissue contrast resolution, but also provides multiparametric information by using a variety of contrast agents, such as tissue-targeted agents,
☆ Disclosure of any conflict of interest: I certify that all authors have had no relevant financial interests or personal affiliations in connection with the content of this manuscript. ⁎ Corresponding author at: Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea. Tel.: +82 2 3410 6438; fax: +82 2 3410 0049. E-mail address: [email protected] (Y.K. Kim). http://dx.doi.org/10.1016/j.mri.2014.08.029 0730-725X/© 2014 Elsevier Inc. All rights reserved.
as well as baseline imaging and novel sequences, such as diffusionweighted imaging (DWI). Gadoxetic acid is a novel dual-acting MR contrast agent for liver imaging that offers the combined properties of an extracellular fluid (ECF) contrast agent during the early vascular–interstitial phases and a liver-specific agent during the hepatobiliary phase (HBP) [3–6]. Inclusion of an HBP benefits clinicians in the diagnosis of HCC, since HCCs are more frequently hypointense during HBP than during the portal venous phase or delayed phase of conventional dynamic CT or MRI [7–11]. However, this benefit might be offset by the fact that other focal liver lesions besides HCC, particularly intrahepatic cholangiocarcinoma (ICC), might show a similar enhancement pattern to HCC due to a short-acting duration of gadoxetic acid as an ECF agent and its early hepatocyte uptake [12–15]. In addition, it is challenging to differentiate HCCs showing iso- or hyperintensity on HBP due to increased uptake of the contrast media (i.e. well- or moderately differentiated HCC) from focal nodular hyperplasia (FNH) [16]. Thus, there is controversy regarding whether gadoxetic acid-enhanced imaging should be introduced into the noninvasive
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
diagnostic criteria for HCC. Due to recent tremendous advances in image quality, DWI is routinely used as a standard clinical liver MR protocol [17–19]. Given that hepatocarcinogenesis is related to increased cellular density, diffusion restriction in the hepatocarcinogenetic pathway is highly indicative of HCC development. In that sense, previous researchers have shown benefit in combining gadoxetic acid and DWI in the detection and characterization of small HCCs that do not fit to HCC criteria on conventional dynamic imaging, as well as the differentiation between HCC and other liver tumors based on signal intensity pattern on HBP or DWI [14,19,20]. To the best of our knowledge, limited research has been conducted to assess the diagnostic performance of gadoxetic acidenhanced MRI protocol including DWI in the diagnosis of focal liver lesions with emphasis on lesion characterization in patients who are at high risk of developing HCC. Since there are pros and cons with regard to the use of gadoxetic acid in HCC workup, assessing the performance of a state of the art MRI protocol using gadoxetic acid in the characterization of focal liver lesions has clinical impact. Therefore, we conducted this study to investigate the preoperative diagnostic performance of a gadoxetic acid-enhanced MRI protocol including DWI with an emphasis on the characterization of hepatic tumors in high-risk HCC patients with chronic liver disease. 2. Materials and methods 2.1. Patients Our institutional review board approved this retrospective study, and informed patient consent was waived. We searched our hospital’s surgical database in the date range between January 2013 and December 2013. This search identified 258 patients who had undergone surgery for hepatic tumors. The inclusion criteria were as follows: (a) patients with chronic hepatitis or cirrhosis who had surgically proven hepatic tumors and (b) patients who underwent liver MRI before surgery. Of the 258 patients, 114 were excluded because of lack of underlying chronic hepatitis or cirrhosis (n = 110) or treatment for HCC prior to an MR examination (n = 4). The final cohort included 144 patients (102 men, 42 women; age range 33–74 years; mean age 57 years) who had a total of 148 HCCs, 13 ICCs, 12 hemangiomas, three hepatocellular adenomas (HCAs), two FNH, one high-grade dysplastic nodule (HGDN), one low-grade dysplastic nodule (LGDN), one large regenerative nodule (RN), one reactive lymphoid hyperplasia (RLH), and one bile duct adenoma (BDA). The causes of chronic liver disease were liver cirrhosis (n = 95) or chronic hepatitis (n = 8) associated with viral hepatitis B, liver cirrhosis (n = 34) or chronic hepatitis (n = 2) associated with viral hepatitis C, and liver cirrhosis associated with both viral hepatitis B and viral hepatitis C (n = 5). Based on the Child–Pugh classification, 140 patients were classified as Child–Pugh class A and four as class B. The diagnosis of all solid hepatic tumors and five hemangiomas was based on a histopathological examination of the surgical specimens. The mean time interval between the MR examination and the surgery was 16 days (range 1–27 days). The operations included segmentectomy (n = 97), lobectomy (n = 45), and liver transplantation (n = 2). The remaining seven hemangiomas were diagnosed based on typical findings, including nodular or globular enhancement with centripetal enhancement patterns on dynamic CT or MRI, very high signal intensity on both moderate and heavily T2-weighted images (T2WI), and stability for at least 12 months of follow-up. 2.2. MR examination MRIs were acquired using a 3.0-T MR system (Intera Achieva 3.0-T, Philips Healthcare, Best, The Netherlands) equipped with a dual-source parallel radiofrequency transmission system and a
1205
quadrature body coil. The baseline MRIs included a T1-weighted turbo field-echo in-phase and opposed sequence (TR/first echo TE, second echo TE, 10/2.3 [in-phase], 3.45 [opposed-phase]; flip angle, 15°; matrix size, 256 × 194; bandwidth, 434.3 Hz/ pixel), a breathhold multishot T2WI with an acceleration factor of 2 (1796/70; flip angle, 90°; matrix size, 324 × 235; bandwidth, 258.4 Hz/pixel), a respiratory-triggered single-shot heavily T2WI with an acceleration factor of 2 (1802/160; flip angle, 90°; matrix size, 252 × 254; bandwidth, 420.9 Hz/pixel) and a 5 mm section thickness, and a field of view of 32–38 cm. Diffusion-weighted single-shot echo planar imaging with the simultaneous use of respiratory triggering was performed using a TR/ TE of 1600/70. The scanning parameters were as follows: b-values of 0, 100 and 800 s/mm2; spectral presaturation with inversion recovery for fat suppression; matrix size, 124 × 124; SENSE acceleration factor, 4.0; field of view, 35 × 35 cm; number of excitations, 3; slice thickness, 5 mm; and slice gap, 1 mm. The apparent diffusion coefficient (ADC) was calculated using a mono-exponential function at b-values of 100 and 800 s/m2. Acquisition time for this sequence was 2–3 min depending on the respiratory efficiency of the patient. Thus, total image acquisition time for precontrast MRI including DWI was 6–7 min. Unenhanced, arterial-phase (20–35 s), portal-phase (60 s), 3 min late-phase, and 20-min hepatobiliary phase (HBP) were obtained for gadoxetic acid-enhanced imaging using a T1-weighted 3D turbofield-echo sequence (enhanced T1 high-resolution isotropic volume examination, eTHRIVE, Philips Healthcare) (3.1/1.5; flip angle, 10°; matrix size, 256 × 256; bandwidth, 724.1 Hz/pixel) with the spectral attenuated inversion recovery fat suppression technique, a 2-mm section thickness, and a field of view of 32–38 cm. The time for the arterial phase imaging was determined using the MR fluoroscopic bolus detection technique. The contrast agent was administered intravenously using a power injector at a rate of 1 mL/s for a dose of 0.025 mmol/kg body weight, followed by a 20-mL saline flush. 2.3. Image analysis Two on-site (observer 1 and 3) and one off-site (observer 2) gastrointestinal radiologists (with 13, five and three years of experience in liver MRI interpretation, respectively) retrospectively reviewed MRIs independently on a picture archiving and communication system (Centricity 3.0, General Electric Medical Systems, Milwaukee, WI). They knew that the patients were at risk for HCC but were blinded to the initial MRI report and pathologic diagnosis of hepatic tumors. Each observer documented the presence of hepatic tumor using a four-point confidence scale (1: probably not a lesion. 2: possibly a lesion, 3: probably a lesion, 4: definitely a lesion) and the possibility of HCC using a four-point confidence scale (1: probably not HCC. 2: possibly HCC, 3: probably HCC, 4: definitely HCC). As for classifying as HCC or not HCC, hepatic nodule considered as DN or hepatic tumors other than HCC was assigned as a confidence level of 1 or 2. Missed lesions were given a rating of 0. Observers were also asked to make a specific diagnosis of detected lesions as HCC, ICC, or other benign lesion such as hemangioma, FNH, or HCA. Observers made a diagnosis of liver tumors based on subjective features of the lesions (including the size, margin, shape, homogeneity, signal intensity, presence of fat, central scar, capsule, and mosaic pattern) and the major features used as diagnostic criteria for HCC (nodule showing enhancement foci during the arterial phases and early washout on portal phase and/or 3 min-late phase, and/or hypointensity on HBP with/without ancillary findings such as hyperintensity on T2WI, mosaic appearance, and tumor capsule), for ICC (nodule showing early peripheral enhancement with a centrally enhanced area and peripheral hypointense rim on 3 min-late phase and or 10 min- and 20 min HBP), for hepatic hemangioma (centripetal enhancement during dynamic MRI and bright hyperintensity similar
1206
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Table 1 Az values and sensitivities obtained with gadoxetic acid-enhanced MRI with diffusion weighted imaging for the detection of 183 hepatic tumors. Az value Observer 1 Observer 2 Observer 3 Pooled data
0.986 0.975 0.971 0.978
± ± ± ±
Sensitivity 0.008 0.012 0.012 0.006
(95% (95% (95% (95%
CI: CI: CI: CI:
0.958–0.998) 0.941–0.992) 0.936–0.990) 962–0.988)
98.4% 97.8% 96.2% 97.5%
(95% (95% (95% (95%
False negative lesions CI:95.3–99.7) CI:94.5–99.4) CI:92.3–98.4) CI:95.8–98.6)
[180] [179] [176] [535]
3 HCCs 2 HCCs, 1 RN, 1 Hemangioma 6 HCCs, 1 RN 6 HCCs, 1 RN, 1 Hemangioma
Note: Data represent Az values ± 1SD; numbers in brackets represent the number of true-positive lesions. HCC, hepatocellular carcinoma; RN, regenerative nodule.
to cysts on T2WI), and for FNH (presence of a central scar, serrated outer contour without capsule, homogeneous arterial hyperenhancement without washout during portal phase, iso- or slightly hypointense compared with the liver on T1WI, iso- or slightly hyperintense compared with the liver on T2WI, and iso-to hyperintensity on HBP) [21–23]. The diagnosis of HCA was based on lesion heterogeneity due to fat component (signal drop on opposed-phase T1WI) or hemorrhage, a strong arterial enhancement, strong hyperintensity on T2WI, and hypointensity on HBP) [23,24]. On DWI, a lesion was considered to be malignant when it was hyperintense at b-100 s/mm2, remained hyperintense at b-800 s/mm2, and showed ADC value that was lower than or equal to that of liver parenchyma [18]. However, since DWI has limited value for characterizing solid liver tumors [25], it was mainly used for lesion localization. We used a target appearance consisting of a central hypointense area and a peripheral hyperintense rim on b-800 DWI as a sign of tumor with abundant fibrosis such as ICC, together with a target appearance consisting of a centrally enhanced area and peripheral hypointense rim on 3 min delayed phase or HBP [14,26]. Since there have been no precise imaging criteria for cirrhosis-associated benign hepatocellular nodules, hepatic nodules that did not fully fit the aforementioned criteria, including lesions assigned as confidence level 1 or 2 during review for lesion detection, were considered to be DN [20]. Although these criteria were presented as rough reference standards to the observers, the final decision was made by the subjective judgment of each observer. To support the results of qualitative analysis, quantitative measurements were performed by one radiologist who did not attend the qualitative analysis by measuring signal intensity (SI) of the liver, HCCs, and ICCs using operator-defined regions of interest (ROIs) in the arterial phase, portal phase, 3 min-late phase, and 10 min- and 20 min HBP. This observer was blinded to the final diagnosis of tumors. For the measurements of SIliver, each ROI area was drawn to avoid the inclusion of confounding structures such as blood vessels or artifacts. Subsequently, liver-to-lesion signal ratio (SR) was calculated as [(SIlesion − SIliver)/SIliver]. In addition, to support target appearance of ICC in qualitative analysis, we measured both signal ratio of central enhancing portion and peripheral rim of ICC on 10 min HBP.
SPSS version 18.0; SPSS, Chicago, IL, USA). An alternative-free response receiver operating characteristic (ROC) curve analysis was performed on a lesion-by-lesion basis for detection and classification of hepatic lesions [27]. The diagnostic accuracy was determined by measuring the area under the ROC curve (Az). A non-parametric analysis of clustered ROC curve data was performed, considering the possible influence of lesion clustering on diagnostic accuracy, using the method proposed by Obuchowski [28]. The sensitivities for tumor detection and classification as HCC or not HCC were calculated according to the number of lesions assigned a confidence level of 3 or 4. The specificity for HCC diagnosis was also calculated. In addition, the number of tumors that were correctly characterized according to the reference standard was calculated for each observer. The κ statistic for multiple observers was calculated to assess interobserver agreement for lesion detection and characterization. A κ value b 0.20 indicated positive but poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; and greater than 0.81, excellent agreement [29].
2.4. Statistical analysis
3.2. Detection of focal liver lesions
Statistical analyses were performed using statistical software (MedCalc version 11.4, MedCalc Software; Mariakerke, Belgium;
The calculated Az value and sensitivities for detecting all lesions are described in Table 1. In pooled data of three observers, Az value and
3. Results 3.1. Histopathologic results Of a total of 144 patients, 123 patients had 148 HCCs (size range: 0.4–10.0 cm; mean: 3.3 ± 2.6 cm), 12 hemangiomas (size range: 0.4–1.5 cm), and one LGDN (1.5 cm). Based on pathological analysis, 100 patients had one solitary HCC each, 21 patients had two HCCs each, and two patients had three HCCs each. Of 148 HCCs, seven lesions were grade I, 128 lesions were grade II, and the remaining 13 lesions were grade III according to Edmondson’s classification of HCC [30]. Nine HCCs were proven to be scirrhous HCC because they had fibrous stroma in ≥ 50% of the tumor area by histological morphometry in the greatest dimension of the cut surface [31]. The remaining 21 patients had 13 ICCs (size range: 1.8–11.0 cm; mean: 4.3 ± 2.5 cm), three HCAs (5.2 cm, 8.0 cm, and 10.5 cm), two FNHs (1.2 and 3.3 cm), one HGDN (0.7 cm), one large RN (0.7 cm), one RLH (1.4 cm), and one BDA (1.0 cm). One patient had both single HGDN and one large RN, and the remaining 20 patients had one lesion each.
Table 2 Values obtained with gadoxetic acid-enhanced MRI with diffusion weighted imaging for the lesion classification as HCC or non-HCC.
Az value Sensitivity Specificity
Observer 1
Observer 2
Observer 3
Pooled data
0.952 ± 0.018 / 95% CI: 0.912–0.978 93.9% (139) / 95% CI: 88.4–97.0 80.0% [7] / 95% CI: 62.5–90.9
0.906 ± 0.027 / 95% CI: 0.856–0.943 91.9% (136) / 95% CI: 86.0–95.5 74.3% [9] / 95% CI: 56.4–86.9
0.910 ± 0.027 / 95% CI: 0.860–0.946 91.9% (136) / 95% CI: 86.0–95.5 68.6% [11] / 95% CI: 50.6–82.6
0.922 ± 0.014 / 95% CI:0.897–0.943 92.6% (411) / 95% CI: 89.6–94.8 74.3 % [27] / 95% CI: 64.7–82.1
Note: Data represent Az values ± 1SD; numbers in parentheses and brackets represent the number of true-positive lesions and false-positive lesions, respectively.
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1207
Table 3 Results of tumor characterization with gadoxetic acid-enhanced imaging with diffusion-weighted imaging by three observers.
Observer 1 Observer 2 Observer 3
Correct diagnosis Misdiagnosis Correct diagnosis Misdiagnosis Correct diagnosis Misdiagnosis
HCC (N = 148)
ICC (N = 13)
Hemangioma (N = 12)
HCA (N = 3)
FNH (N = 2)
RLH (N = 1)
BDA (N = 1)
DN or RN (N = 3)
139 DN (5) ICC (1) Missed (3) 136 DN (9) ICC (1) Missed (2) 136 DN (5) ICC (2) Missed (5)
12 HCC (1) 11 HCC (2) 10 HCC (3)
11 HCC (1) 11 Missed (1) 12 0
2 HCC (1) 1 HCC (2) 0 HCC (3)
1 HCC (1) 0 HCC (2) 0 HCC (2)
0 HCC (1) 0 HCC (1) 0 HCC (1)
0 HCC (1) 0 HCC (1) 0 HCC (1)
2 HCC (1) 1 HCC (1) Missed (1) 1 HCC (1) Missed (1)
Note: Data represent the number of tumors. HCC, hepatocellular carcinoma; ICC, intrahepatic cholangiocarcinoma; HCA, hepatocellular adenoma; FNH, focal nodular hyperplasia; RLH, reactive lymphoid hyperplasia; BDA, bile duct adenoma; DN, dysplastic nodule; RN, regenerative nodule.
sensitivity are 0.978 ± 0.006 and 97.5%, respectively (Table 1). The lesions (size range: 0.2–0.7 cm) that were not verified on MRI was three HCCs in observer 1; two HCCs, one large RN, and one hemangioma in observer 2; and six HCCs and one large RN in observer 3. All HCCs not verified by reviewers were daughter nodules or coexisting tiny HCCs of the main tumor. On reviewing DWI, 144 HCCs
except four and all 13 ICCs showed hyperintensity. Of solid benign lesions, DWI hyperintensity was observed in 3 HCAs, 1 FNH, 1 RLH, 1 BDA, and 1 DN. There were two HCCs (0.4 cm and 0.6 cm) in two patients that were not verified by any observer (confidence rating 0). On review, one was not discernible even on retrospective review, and one was seen as faintly hypointense on HBP.
Fig. 1. Surgically confirmed 1.5-cm-sized hepatocellular carcinoma in a 61-year-old man. Axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) 3 min delay, and (D) 20 min HBP after administration of gadoxetic acid show a small hepatic mass with arterial hypervascularization (arrow) and delayed hypointensity (arrows). This tumor is clearly seen as a hyperintense and dark area (D) on single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2 and (E) on the ADC map (arrows), respectively. (F) Gross specimen confirmed a small hepatocellular carcinoma.
1208
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Fig. 2. Surgically confirmed 1.8-cm-sized intrahepatic cholangiocarcinoma in a 65-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) 3 min delay, and (C) 10 min HBP image after administration of gadoxetic acid, the tumor shows arterial hypervascularization (arrow) and delayed hypointensity (arrows). The faint hyperintense area in the center is suspicious on the 10 min HBP phase (C) but is not definitive. On single-shot echo-planar diffusion weighed imaging (D) at b = 800 s/mm2, the tumor shows clear hyperintensity with a faint hypointense area in the center (arrow). The tumor was misdiagnosed as HCC by all observers. (E) Gross specimen confirmed a small cholangiocarcinoma.
3.3. Characterization of focal liver lesions The calculated Az value and sensitivities for classifying tumors as HCC or not HCC are described in Table 2. Az values were 0.952 ± 0.018 for observer 1, 0.906 ± 0.027 for observer 2, and 0.910 ± 0.027 for observer 3, with 0.922 ± 0.014 for pooled data. Among 183 focal liver lesions, 91.3% (167/183, observer 1), 87.4% (160/183, observer 2), and 86.9% (159/183, observer 3) were correctly characterized according to the reference standard, and 89.7% of pooled data were correctly characterized (Table 3). On reviewing 148 HCCs, 140 HCCs showed arterial hyperenhancement, hypointensity on HBP, and hyperintensity on DWI (Fig. 1) [6–11]. On 3 min late phase, 137 of these 140 HCCs showed a variable degree of hypointensity, and the remaining three were isointense to liver parenchyma. Other than two to five tiny HCCs missed by each observer, five or nine HCCs (0.4–1.3 cm in diameter) were misclassified as non-HCC (DN) by each observer. On retrospective review, most of them did not clearly show arterial hyperenhance-
ment or hypointensity on late phase or HBP or hyperintensity on DWI. One scirrhous HCC was interpreted as ICC by all observers. Thus, observer 1 correctly characterized 93.9% (n = 139) of HCCs, and observers 2 and 3 correctly characterized 91.9% (n = 136) of HCCs, respectively. Specificities for HCC diagnosis were 80.0% for observer 1, 74.3% for observer 2, and 68.6% for observer 3, with a 74.3% specificity for pooled data. Of 13 ICCs, 12, 11, and 10 were correctly characterized by each observer, respectively, and between one and three (1.8–2.8 cm in diameter) were misinterpreted as HCC by each observer (Fig. 2). On reviewing 3 ICCs misinterpreted, they were observed as arterially hyperenhancing mass and showed faint central enhancement with hypointense rim on 10 min HBP and/or hyperintensity with central hypointense area on b-800 DWI. Six ICCs were considered to be hypervascular tumors because they showed arterial hyperenhancement in more than 50% of the total tumor area (Fig. 2). On 3 min late phase, two tumors were seen as mostly isointense to the liver, and four were seen as hypointense with variable degrees of central enhancing
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1209
Fig. 3. Surgically confirmed 2.0-cm-sized intrahepatic cholangiocarcinoma in a 57-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) portal venous phase, and (C) 10 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows arterial hypervascularization and hypointensity on B and C. On 10-min HBP, the tumor (arrow) shows a target appearance consisting of central enhancement and a peripheral hypointense rim. (D) On single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2, the tumor also shows a target appearance revealing peripheral hyperintensity compared to the central portion (arrow).
area. Twelve ICCs showed a centrally enhanced area and peripheral hypointense rim (target appearance) on 10 min HBP as well as a target appearance consisting of a central hypointense area and a peripheral hyperintense rim on b-800 DWI (Fig. 3). The difference in enhancement features between HCC and ICC could be supported by the results of quantitative analysis (Table 4) as HCC showed higher mean values of SR on arterial phase and portal phase, and thereafter lower SR than ICC. On 10 min HBP, the central area of ICC tended to show higher SR than tumor periphery (mean value; −0.222 ± 0.181 vs −0.341 ± 0.176), indicating target appearing enhancement. With regard to the characterization of the three HCAs and two FNHs, observer 1 misinterpreted one HCA and one FNH as HCCs. Observers 2 and 3 misinterpreted two FNHs (Fig. 5) and two or three HCAs as being HCCs. These misinterpreted lesions showed arterial hyperenhancement and hypointensity on HBP and/or signal decrease on the opposed phase compared to an in-phase chemical shift imaging indicating intralesional fat (Fig. 4). All FNHs had no central scar. One (1.2 cm) fitted HCC imaging criteria, but showed hyperintense rim on HBP. Another FNH (3.0 cm) showed no arterial hyperenhancement and hypointensity on HBP with isointensity on T1- and T2WI. RLH, BDA, and one HGDN were all misinterpreted as HCCs by all three
observers because these lesions showed arterial hyperenhancement, hypointensity on HBP, and hyperintensity on DWI. 3.4. Inter-observer agreement The kappa values were 0.713 (observer 1 vs. 2), 0.622 (observer 1 vs. 3), and 0.692 (observer 2 vs. 3) for tumor detection and 0.743 (observer 1 vs. 2), 0.743 (observer 1 vs. 3), and 0.749 (observer 2 vs. 3) for tumor characterization, thus indicating good inter-observer agreement. 4. Discussion This study was conducted in a viral hepatitis endemic population with a high risk of developing HCC. Owing to early hepatocyte uptake of gadoxetic acid, the majority of focal liver lesions, excluding lesions that preserve hepatocyte function, tend to show hypointensity on HBP and even on 3 min late phase, which might increase the chance of misdiagnosing a hypervascular lesion as HCC in a viral hepatitis endemic population. Apart from these considerations, the majority of benign tumors included were surgically resected, which possibly explains their atypical imaging features similar to those of
Table 4 Results of signal ratio of hepatocellular carcinoma and intrahepatic cholangiocarcinoma on each phase.
HCC ICC
Arterial phase
Portal phase
3 min late phase
10 min HBP
20 min HBP
0.588 ± 0.371 0.081 ± 0.299
−0.077 ± 0.098 −0.097 ± 0.169
−0.217 ± 0.141 −0.113 ± 0.139
−0.298 ± 0.196 −0.258 ± 0.170
−0.464 ± 0.074 −0.353 ± 0.151
Note: Data represent mean values ± SD. HCC, hepatocellular carcinoma; ICC, intrahepatic cholangiocarcinoma; HBP, hepatobiliary phase.
1210
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Fig. 4. Surgically confirmed 5.2-cm-sized hepatocellular adenoma in a 33-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase and (B) 20 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows arterial hypervascularization and hypointensity, respectively. (C) Opposed-phase spoiled gradient-echo MRI shows signal drop compared to the (D) in-phase image, indicating diffuse fatty infiltration of the tumor (arrow). The tumor was misdiagnosed as HCC by all observers. (E) Gross specimen confirmed a hepatocellular adenoma with fat.
hepatic malignancies. Thus, biases could have been introduced during image review or population selection. Given that lesion characterization is made not only by enhancement pattern, but also by ancillary findings on each MR sequence in daily practice, we hypothesized that a combined reading of whole MR sequences would offset such limitations and bias. With regard to tumor detection, given that the gadoxetic acid MRI protocol yielded high sensitivity (97.5% of pooled data) and accuracy (0.978 of pooled data) for detecting 183 hepatic tumors, our study reaffirmed previous studies showing promising sensitivity of the combination of gadoxetic acid MRI and DWI for detecting HCC [19,20]. This might be attributable to a mixed contribution of HBP and DWI, which offer excellent lesion-to-liver contrast. Other than two to five tiny HCCs missed by all observers, all five or nine HCCs misclassified as DN by the individual observers were daughter nodules or satellite nodules with non-clear signal change on arterial phase or HBP and DWI.
Since liver transplantation is an established treatment option for HCC, but not for ICC, precise differentiation between ICC and HCC is of paramount importance. A major concern regarding use of gadoxetic acid in HCC workup is misdiagnosis of ICC as HCC due to pseudowashout phenomenon on 3 min late phase [15]. This is particularly true for small ICC due to its tendency to demonstrate arterial hypervascularity [12,14,15]. On 3 min late phase image, four of six hypervascular ICCs were hypointense and two were isointense, whereas three HCCs were isointense. Given that small HCCs frequently show no contrast washout on portal venous or 3 min late phase images after administration of an ECF agent, as ICC does [6–9], the differentiation between HCC and ICC does not highly rely on whether tumoral enhancement is sustained on delayed phase. In practice, even when using an ECF agent, there may be uncertainty in distinguishing the two tumors. In this study, of 13 ICCs, between one and three small hypervascular ICCs (1.8–2.8 cm in diameter) were misdiagnosed as HCC by each observer. On review, three ICCs showed faint central
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1211
Fig. 5. Surgically confirmed 1.2-cm-sized focal nodular hyperplasia in a 63-year-old woman. On axial 3D T1-weighted GRE images obtained at (A) arterial phase and (B) 20 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows hyperenhancement and hypointensity, respectively. However, the tumor shows a hyperintense rim at 20 min HBP. (C) On single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2, the tumor is clearly seen as hyperintense. The tumor was misdiagnosed as HCC by two observers. (D) Gross specimen confirmed a foal nodular hyperplasia.
enhancement with a hypointense rim on 10 min HBP and/or hyperintensity with central hypointense area on b-800 DWI, both of which are considered indicators of tumor with abundant fibrosis such as ICC [14,15,26]. Although the majority of ICCs were seen as mostly hypointense on 3 min late phase and/or HBP, similar to HCCs, all ICCs except one demonstrated a target appearance on 10 min HBP and b-800 DWI, suggesting ICC with fibrous stroma [14,15,26]. This was supported by the results of quantitative analysis. Thus, it is necessary for radiologists to discern the aforementioned features for accurate characterization of ICC. In addition, most ICCs tended to show stronger enhancement in the tumor periphery than in the center, which might be responsible for the correct characterization of seven scirrhous HCCs except two as HCC by all observers, although similar features in HBP and DWI might be observed in both ICC and scirrhous HCC due to their fibrotic components. For the diagnosis of FNH and HCA with standard MRI, presence of central scar or capsule, intralesional fat, tumor contour, signal intensity on T1- and T2WI as well as dynamic enhancement pattern were all considered [21–24]. However, based on findings on standard MRI, 40% of FNHs and HCAs were inconclusive because of the lack of their typical features [22]. In such case, HBP is reportedly useful to differentiate between FNH and HCA because most FNHs appear at least partially hyperintense to the liver, while most HCAs are seen as hypointense on HBP [21,22]. However, two FNHs were seen as primarily hypointense on HBP and were misdiagnosed as HCC. On review, one (1.2 cm) showed a hyperintense rim on HBP, which might be an important sign indicating FNH (Fig. 5) [21,32]. Another FNH (3.0 cm) showed neither arterial hyperenhancement nor enhancement on HBP, but displayed a serrated outer contour, which could be a differentiating feature from encapsulated HCC or HCA. Since HCA might show arterial hyperenhancement, HBP hypointensity, and a fat component, misdiagnosis of HCA as HCC is possible, as was shown in our cases. However, despite of large tumor size, HCA did not show a mosaic appearance with intratumoral septum, which is frequently seen in advanced HCC
[33,34]. Based on the above considerations, discerning ancillary features of a tumor on MRI, which is not afforded by other imaging modalities, makes accurate tumor characterization possible. RLH and BDA showed similar features to HCC, revealing the need for further large-scale evaluations to determine specific imaging features. The imaging differentiation between HCC and DN is also challenging. One HGDN was misdiagnosed as HCC by all observers. Even on review, it was compatible with hypervascular HCC, indicating incomplete differentiation between HCC and DN with the application of any imaging modality. Our study had limitations. First, it included only patients with chronic liver disease who underwent surgery, which might have caused a selection bias. In particular, the proportion of hypervascular ICCs was higher than previous studies [14,15,26]. This may be due to tumor surveillance in patients with chronic liver disease, which could lead to early detection of small ICCs that are more frequently seen as hypervascular mass than larger ICCs. However, our study did not primarily focus on tumor incidence or proportion of specific tumor type. Benign solid tumors included were all surgically resected, which may have caused an underestimation of the ability of MRI to characterize tumors. Second, issues regarding the differentiation between HCC and DN were not fully addressed because only two DNs and one large RN were macroscopically detectable. In addition, it is not possible to recruit all cirrhosisassociated benign hepatocellular nodules that were only pathologically proven but not verified on imaging and vice versa. Third, intraindividual comparison between current MRI protocol and MRI with conventional ECF agent was not made, creating uncertainty regarding the relative efficacy of the current MRI protocol for tumor characterization in chronic liver disease. In addition, a similar study using a similar MR imager and ECF MR contrast agent has not been conducted. Fourth, the result might not reflect the performance of only gadoxetic acid-enhanced MRI because reviewers were allowed to read whole MR sequences to approximate real practice.
1212
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
In conclusion, despite concerns regarding limitations in tumor characterization with gadoxetic acid-enhanced MRI due to its early hepatocyte uptake, gadoxetic acid-enhanced MRI including DWI showed a reasonable performance for tumor characterization by offering high sensitivity for tumor detection in patients with chronic hepatitis or cirrhosis. It is necessary for radiologists to discern ancillary features beyond the enhancement characteristics of variable hepatic tumors provided by MRI in order to better characterize hepatic tumors.
References [1] Yuen MF, Cheng CC, Lauder IJ, Lam SK, Ooi CG, Lai CL. Early detection of hepatocellular carcinoma increases the chance of treatment: Hong Kong experience. Hepatology 2000;31:330–5. [2] Bruix J, Llovet JM. Major achievements in hepatocellular carcinoma. Lancet 2009; 373:614–6. [3] Haradome H, Grazioli L, Tinti R, et al. Additional value of gadoxetic acid-DTPAenhanced hepatobiliary phase MR imaging in the diagnosis of early-stage hepatocellular carcinoma: comparison with dynamic triple-phase multidetector CT imaging. J Magn Reson Imaging 2011;34:69–78. [4] Jung G, Breuer J, Poll LW, et al. Imaging characteristics of hepatocellular carcinoma using the hepatobiliary contrast agent GD-EOB-DTPA. Acta Radiol 2006;47:15–23. [5] Brismar TB, Dahlstrom N, Edsborg N, et al. Liver vessel enhancement by Gd-BOPTA and Gd-EOB-DTPA: a comparison in healthy volunteers. Acta Radiol 2009;50:709–15. [6] Park G, Kim YK, Kim CS, et al. Diagnostic efficacy of gadoxetic acid-enhanced MRI in the detection of hepatocellular carcinomas: comparison with gadopentetate dimeglumine. Br J Radiol 2010;83:1010–6. [7] Sun HY, Lee JM, Shin CI, et al. Gadoxetic acid-enhanced magnetic resonance imaging for differentiating small hepatocellular carcinomas (b=2 cm in diameter) from arterial enhancing pseudolesions: special emphasis on hepatobiliary phase imaging. Invest Radiol 2010;45:96–103. [8] Kim YK, Kim CS, Han YM, et al. Detection of small hepatocellular carcinoma: can gadoxetic acid-enhanced magnetic resonance imaging replace combining gadopentetate dimeglumine-enhanced and superparamagnetic iron oxideenhanced magnetic resonance imaging? Invest Radiol 2010;45:740–6. [9] Ahn SS, Kim MJ, Lim JS, et al. Added value of gadoxetic acid-enhanced hepatobiliary phase MR imaging in the diagnosis of hepatocellular carcinoma. Radiology 2010;255:459–66. [10] Kim YK, Kim CS, Han YM, et al. Detection of hepatocellular carcinoma: gadoxetic acid-enhanced 3-dimensional magnetic resonance imaging versus multidetector row computed tomography. J Comput Assist Tomogr 2009;33:844–50. [11] Kim SH, Kim SH, Lee J, et al. Gadoxetic acid-enhanced MRI versus triple-phase MDCT for the preoperative detection of hepatocellular carcinoma. Am J Roentgenol 2009;192:1675–81. [12] Kim SA, Lee JM, Lee KB, et al. Intrahepatic mass-forming cholangiocarcinomas: enhancement patterns at multiphasic CT, with special emphasis on arterial enhancement pattern-correlation with clinicopathologic findings. Radiology 2011;260:148–57. [13] Kim SJ, Lee JM, Han JK, Kim KH, Lee JY, Choi BI. Peripheral mass-forming cholangiocarcinoma in cirrhotic liver. AJR Am J Roentgenol 2007;189:1428–34. [14] Park HJ, Kim YK, Park MJ, Lee WJ. Small intrahepatic mass-forming cholangiocarcinoma: target sign on diffusion-weighted imaging for differentiation from hepatocellular carcinoma. Abdom Imaging 2013;38:793–801.
[15] Jeong HT, Kim MJ, Chung YE, Choi JY, Park YN, Kim KW. Gadoxetate disodiumenhanced MRI of mass-forming intrahepatic cholangiocarcinomas: imaginghistologic correlation. AJR Am J Roentgenol 2013;201:W603–11. [16] Kitao A, Zen Y, Matsui O, et al. Hepatocellular carcinoma: signal intensity at gadoxetic acid-enhanced MR imaging-correlation with molecular transporters and histopathologic features. Radiology 2010;256:817–26. [17] Taouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254:47–66. [18] Nasu K, Kuroki Y, Nawano S, et al. Hepatic metastases: diffusion-weighted sensitivity-encoding versus SPIO-enhanced MR imaging. Radiology 2006;239: 122–30. [19] Park MJ, Kim YK, Lee MW, et al. Small hepatocellular carcinomas: improved sensitivity by combining gadoxetic acid-enhanced and diffusion-weighted MR imaging patterns. Radiology 2012;265:104–14. [20] Park MJ, Kim YK, Lee MH, Lee JH. Validation of diagnostic criteria using gadoxetic acid-enhanced and diffusion-weighted MR imaging for small hepatocelllular carcinoma (b=2.0 cm) in patients with hepatitis-induced liver cirrhosis. Acta Radiol 2013;54:127–36. [21] Grieser C, Steffen IG, Seehofer D, et al. Histopathologically confirmed focal nodular hyperplasia of the liver: gadoxetic acid-enhanced MRI characteristics. Magn Reson Imaging 2013;31:755–60. [22] Bieze M, van den Esschert JW, Nio CY, et al. Diagnostic accuracy of MRI in differentiating hepatocellular adenoma from focal nodular hyperplasia: prospective study of the additional value of gadoxetate disodium. AJR Am J Roentgenol 2012;199:26–34. [23] Ronot M, Vilgrain V. Imaging of benign hepatocellular lesions: current concepts and recent updates. Clin Res Hepatol Gastroenterol 2014 [in press]. [24] Laumonier H, Bioulac-Sage P, Laurent C, Zucman-Rossi J, Balabaud C, Trillaud H. Hepatocellular adenomas: magnetic resonance imaging features as a function of molecular pathological classification. Hepatology 2008;48:808–18. [25] Taouli B. Diffusion-weighted MR, imaging for liver lesion characterization: a critical look. Radiology 2012;262:378–80. [26] Kim MY, Kim YK, Park HJ, Park MJ, Lee WJ, Choi D. Diagnosis of focal liver lesions with gadoxetic acid-enhanced MRI: is a shortened delay time possible by adding diffusion-weighted imaging? J Magn Reson Imaging 2014;39:31–41. [27] Chakraborty DP, Winter LH. Free-response methodology: alternate analysis and a new observer-performance experiment. Radiology 1990;174:873–81. [28] Obuchowski NA. Nonparametric analysis of clustered ROC curve data. Biometrics 1997;53:567–78. [29] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–74. [30] Edmondson HA, Steiner PE. Primary carcinoma of liver: study of 100 cases among 48,900 necropsies. Cancer 1954;7:462–503. [31] Matsuura S, Aishima S, Taguchi K, et al. 'Scirrhous' type hepatocellular carcinomas: a special reference to expression of cytokeratin 7 and hepatocyte paraffin 1. Histopathology 2005;47:382–90. [32] Fujiwara H, Sekine S, Onaya H, Shimada K, Mikata R, Arai Y. Ring-like enhancement of focal nodular hyperplasia with hepatobiliary-phase Gd-EOBDTPA-enhanced magnetic resonance imaging: radiological–pathological correlation. Jpn J Radiol 2011;29:739–43. [33] Stevens WR, Gulino SP, Batts KP, Stephens DH, Johnson CD. Mosaic pattern of hepatocellular carcinoma: histologic basis for a characteristic CT appearance. J Comput Assist Tomogr 1996;20:337–42. [34] Willatt JM, Hussain HK, Adusumilli S, Marrero JA. MR imaging of hepatocellular carcinoma in the cirrhotic liver: challenges and controversies. Radiology 2008; 247:311–30.
Contents lists available at ScienceDirect
Magnetic Resonance Imaging journal homepage: www.mrijournal.com
Is gadoxetic acid-enhanced MRI limited in tumor characterization for patients with chronic liver disease?☆ Soyi Kwon a, Young Kon Kim a,⁎, Hyun Jeong Park b, Woo Kyoung Jeong a, Won Jae Lee a, Dongil Choi a a b
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea Department of Radiology, Chung-Ang University Hospital, Seoul, Republic of Korea
a r t i c l e
i n f o
Article history: Received 7 April 2014 Revised 9 June 2014 Accepted 10 August 2014 Keywords: Hepatocellular carcinoma Gd-EOB-DTPA Diffusion-weighted imaging Tumor characterization Cholangiocarcinoma
a b s t r a c t Purpose: There are pros and cons to the use of gadoxetic acid in hepatocellular carcinoma (HCC) workup due to the potential for high false positive diagnosis. This study was conducted to investigate the preoperative diagnostic performance of gadoxetic acid-enhanced MRI protocol including diffusionweighted imaging (DWI) with emphasis on tumor characterization developed in high risk HCC patients. Materials and methods: We included 144 patients (102 men, 42 women; age range 33–74 years) with chronic viral hepatitis or cirrhosis and 183 focal hepatic tumors (size range, 0.4–11.0 cm; mean, 3.2 cm), including 148 HCCs, 13 cholangiocarcinomas, 12 hemangiomas, three hepatocellular adenomas, two focal nodular hyperplasias, and five other tumors. All patients underwent gadoxetic acid-enhanced MRI protocol with DWI. MRIs were independently interpreted by three observers for the detection and characterization of hepatic tumors. Results: Sensitivities for detecting all 183 liver tumors were 98.4%, 97.8%, and 96.2% for each observer, respectively, with a 97.5% for pooled data. Among 183 hepatic tumors, 91.3% (n = 167), 87.4% (n = 160), and 86.9% (n = 159) were correctly characterized according to their reference standard by each observer, respectively. In 13 cholangiocarcinomas, one to three were misinterpreted as HCC, and the remaining tumors were correctly characterized by each observer. The accuracies (Az) of MRI for HCC diagnosis were 0.952 for observer 1, 0.906 for observer 2, and 0.910 for observer 3, with 0.922 for pooled data. There was good inter-observer agreement. Conclusion: The gadoxetic acid-enhanced MRI including DWI showed a reasonable performance for tumor characterization with high sensitivity for tumor detection in patients with chronic liver disease, despite concerns of high false positive diagnosis of hypervascular tumors. © 2014 Elsevier Inc. All rights reserved.
1. Introduction The goal of liver imaging for patients with chronic liver disease is the early detection and accurate characterization of hepatocellular carcinoma (HCC) by differentiating it from benign cirrhosisassociated hepatocellular nodules as well as other hepatic tumors. This increases the success rate of curative treatment and leads to a better outcome [1,2]. Magnetic resonance imaging (MRI) has the potential to better fulfill these imaging requirements than other liver imaging modalities since it not only offers excellent soft tissue contrast resolution, but also provides multiparametric information by using a variety of contrast agents, such as tissue-targeted agents,
☆ Disclosure of any conflict of interest: I certify that all authors have had no relevant financial interests or personal affiliations in connection with the content of this manuscript. ⁎ Corresponding author at: Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea. Tel.: +82 2 3410 6438; fax: +82 2 3410 0049. E-mail address: [email protected] (Y.K. Kim). http://dx.doi.org/10.1016/j.mri.2014.08.029 0730-725X/© 2014 Elsevier Inc. All rights reserved.
as well as baseline imaging and novel sequences, such as diffusionweighted imaging (DWI). Gadoxetic acid is a novel dual-acting MR contrast agent for liver imaging that offers the combined properties of an extracellular fluid (ECF) contrast agent during the early vascular–interstitial phases and a liver-specific agent during the hepatobiliary phase (HBP) [3–6]. Inclusion of an HBP benefits clinicians in the diagnosis of HCC, since HCCs are more frequently hypointense during HBP than during the portal venous phase or delayed phase of conventional dynamic CT or MRI [7–11]. However, this benefit might be offset by the fact that other focal liver lesions besides HCC, particularly intrahepatic cholangiocarcinoma (ICC), might show a similar enhancement pattern to HCC due to a short-acting duration of gadoxetic acid as an ECF agent and its early hepatocyte uptake [12–15]. In addition, it is challenging to differentiate HCCs showing iso- or hyperintensity on HBP due to increased uptake of the contrast media (i.e. well- or moderately differentiated HCC) from focal nodular hyperplasia (FNH) [16]. Thus, there is controversy regarding whether gadoxetic acid-enhanced imaging should be introduced into the noninvasive
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
diagnostic criteria for HCC. Due to recent tremendous advances in image quality, DWI is routinely used as a standard clinical liver MR protocol [17–19]. Given that hepatocarcinogenesis is related to increased cellular density, diffusion restriction in the hepatocarcinogenetic pathway is highly indicative of HCC development. In that sense, previous researchers have shown benefit in combining gadoxetic acid and DWI in the detection and characterization of small HCCs that do not fit to HCC criteria on conventional dynamic imaging, as well as the differentiation between HCC and other liver tumors based on signal intensity pattern on HBP or DWI [14,19,20]. To the best of our knowledge, limited research has been conducted to assess the diagnostic performance of gadoxetic acidenhanced MRI protocol including DWI in the diagnosis of focal liver lesions with emphasis on lesion characterization in patients who are at high risk of developing HCC. Since there are pros and cons with regard to the use of gadoxetic acid in HCC workup, assessing the performance of a state of the art MRI protocol using gadoxetic acid in the characterization of focal liver lesions has clinical impact. Therefore, we conducted this study to investigate the preoperative diagnostic performance of a gadoxetic acid-enhanced MRI protocol including DWI with an emphasis on the characterization of hepatic tumors in high-risk HCC patients with chronic liver disease. 2. Materials and methods 2.1. Patients Our institutional review board approved this retrospective study, and informed patient consent was waived. We searched our hospital’s surgical database in the date range between January 2013 and December 2013. This search identified 258 patients who had undergone surgery for hepatic tumors. The inclusion criteria were as follows: (a) patients with chronic hepatitis or cirrhosis who had surgically proven hepatic tumors and (b) patients who underwent liver MRI before surgery. Of the 258 patients, 114 were excluded because of lack of underlying chronic hepatitis or cirrhosis (n = 110) or treatment for HCC prior to an MR examination (n = 4). The final cohort included 144 patients (102 men, 42 women; age range 33–74 years; mean age 57 years) who had a total of 148 HCCs, 13 ICCs, 12 hemangiomas, three hepatocellular adenomas (HCAs), two FNH, one high-grade dysplastic nodule (HGDN), one low-grade dysplastic nodule (LGDN), one large regenerative nodule (RN), one reactive lymphoid hyperplasia (RLH), and one bile duct adenoma (BDA). The causes of chronic liver disease were liver cirrhosis (n = 95) or chronic hepatitis (n = 8) associated with viral hepatitis B, liver cirrhosis (n = 34) or chronic hepatitis (n = 2) associated with viral hepatitis C, and liver cirrhosis associated with both viral hepatitis B and viral hepatitis C (n = 5). Based on the Child–Pugh classification, 140 patients were classified as Child–Pugh class A and four as class B. The diagnosis of all solid hepatic tumors and five hemangiomas was based on a histopathological examination of the surgical specimens. The mean time interval between the MR examination and the surgery was 16 days (range 1–27 days). The operations included segmentectomy (n = 97), lobectomy (n = 45), and liver transplantation (n = 2). The remaining seven hemangiomas were diagnosed based on typical findings, including nodular or globular enhancement with centripetal enhancement patterns on dynamic CT or MRI, very high signal intensity on both moderate and heavily T2-weighted images (T2WI), and stability for at least 12 months of follow-up. 2.2. MR examination MRIs were acquired using a 3.0-T MR system (Intera Achieva 3.0-T, Philips Healthcare, Best, The Netherlands) equipped with a dual-source parallel radiofrequency transmission system and a
1205
quadrature body coil. The baseline MRIs included a T1-weighted turbo field-echo in-phase and opposed sequence (TR/first echo TE, second echo TE, 10/2.3 [in-phase], 3.45 [opposed-phase]; flip angle, 15°; matrix size, 256 × 194; bandwidth, 434.3 Hz/ pixel), a breathhold multishot T2WI with an acceleration factor of 2 (1796/70; flip angle, 90°; matrix size, 324 × 235; bandwidth, 258.4 Hz/pixel), a respiratory-triggered single-shot heavily T2WI with an acceleration factor of 2 (1802/160; flip angle, 90°; matrix size, 252 × 254; bandwidth, 420.9 Hz/pixel) and a 5 mm section thickness, and a field of view of 32–38 cm. Diffusion-weighted single-shot echo planar imaging with the simultaneous use of respiratory triggering was performed using a TR/ TE of 1600/70. The scanning parameters were as follows: b-values of 0, 100 and 800 s/mm2; spectral presaturation with inversion recovery for fat suppression; matrix size, 124 × 124; SENSE acceleration factor, 4.0; field of view, 35 × 35 cm; number of excitations, 3; slice thickness, 5 mm; and slice gap, 1 mm. The apparent diffusion coefficient (ADC) was calculated using a mono-exponential function at b-values of 100 and 800 s/m2. Acquisition time for this sequence was 2–3 min depending on the respiratory efficiency of the patient. Thus, total image acquisition time for precontrast MRI including DWI was 6–7 min. Unenhanced, arterial-phase (20–35 s), portal-phase (60 s), 3 min late-phase, and 20-min hepatobiliary phase (HBP) were obtained for gadoxetic acid-enhanced imaging using a T1-weighted 3D turbofield-echo sequence (enhanced T1 high-resolution isotropic volume examination, eTHRIVE, Philips Healthcare) (3.1/1.5; flip angle, 10°; matrix size, 256 × 256; bandwidth, 724.1 Hz/pixel) with the spectral attenuated inversion recovery fat suppression technique, a 2-mm section thickness, and a field of view of 32–38 cm. The time for the arterial phase imaging was determined using the MR fluoroscopic bolus detection technique. The contrast agent was administered intravenously using a power injector at a rate of 1 mL/s for a dose of 0.025 mmol/kg body weight, followed by a 20-mL saline flush. 2.3. Image analysis Two on-site (observer 1 and 3) and one off-site (observer 2) gastrointestinal radiologists (with 13, five and three years of experience in liver MRI interpretation, respectively) retrospectively reviewed MRIs independently on a picture archiving and communication system (Centricity 3.0, General Electric Medical Systems, Milwaukee, WI). They knew that the patients were at risk for HCC but were blinded to the initial MRI report and pathologic diagnosis of hepatic tumors. Each observer documented the presence of hepatic tumor using a four-point confidence scale (1: probably not a lesion. 2: possibly a lesion, 3: probably a lesion, 4: definitely a lesion) and the possibility of HCC using a four-point confidence scale (1: probably not HCC. 2: possibly HCC, 3: probably HCC, 4: definitely HCC). As for classifying as HCC or not HCC, hepatic nodule considered as DN or hepatic tumors other than HCC was assigned as a confidence level of 1 or 2. Missed lesions were given a rating of 0. Observers were also asked to make a specific diagnosis of detected lesions as HCC, ICC, or other benign lesion such as hemangioma, FNH, or HCA. Observers made a diagnosis of liver tumors based on subjective features of the lesions (including the size, margin, shape, homogeneity, signal intensity, presence of fat, central scar, capsule, and mosaic pattern) and the major features used as diagnostic criteria for HCC (nodule showing enhancement foci during the arterial phases and early washout on portal phase and/or 3 min-late phase, and/or hypointensity on HBP with/without ancillary findings such as hyperintensity on T2WI, mosaic appearance, and tumor capsule), for ICC (nodule showing early peripheral enhancement with a centrally enhanced area and peripheral hypointense rim on 3 min-late phase and or 10 min- and 20 min HBP), for hepatic hemangioma (centripetal enhancement during dynamic MRI and bright hyperintensity similar
1206
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Table 1 Az values and sensitivities obtained with gadoxetic acid-enhanced MRI with diffusion weighted imaging for the detection of 183 hepatic tumors. Az value Observer 1 Observer 2 Observer 3 Pooled data
0.986 0.975 0.971 0.978
± ± ± ±
Sensitivity 0.008 0.012 0.012 0.006
(95% (95% (95% (95%
CI: CI: CI: CI:
0.958–0.998) 0.941–0.992) 0.936–0.990) 962–0.988)
98.4% 97.8% 96.2% 97.5%
(95% (95% (95% (95%
False negative lesions CI:95.3–99.7) CI:94.5–99.4) CI:92.3–98.4) CI:95.8–98.6)
[180] [179] [176] [535]
3 HCCs 2 HCCs, 1 RN, 1 Hemangioma 6 HCCs, 1 RN 6 HCCs, 1 RN, 1 Hemangioma
Note: Data represent Az values ± 1SD; numbers in brackets represent the number of true-positive lesions. HCC, hepatocellular carcinoma; RN, regenerative nodule.
to cysts on T2WI), and for FNH (presence of a central scar, serrated outer contour without capsule, homogeneous arterial hyperenhancement without washout during portal phase, iso- or slightly hypointense compared with the liver on T1WI, iso- or slightly hyperintense compared with the liver on T2WI, and iso-to hyperintensity on HBP) [21–23]. The diagnosis of HCA was based on lesion heterogeneity due to fat component (signal drop on opposed-phase T1WI) or hemorrhage, a strong arterial enhancement, strong hyperintensity on T2WI, and hypointensity on HBP) [23,24]. On DWI, a lesion was considered to be malignant when it was hyperintense at b-100 s/mm2, remained hyperintense at b-800 s/mm2, and showed ADC value that was lower than or equal to that of liver parenchyma [18]. However, since DWI has limited value for characterizing solid liver tumors [25], it was mainly used for lesion localization. We used a target appearance consisting of a central hypointense area and a peripheral hyperintense rim on b-800 DWI as a sign of tumor with abundant fibrosis such as ICC, together with a target appearance consisting of a centrally enhanced area and peripheral hypointense rim on 3 min delayed phase or HBP [14,26]. Since there have been no precise imaging criteria for cirrhosis-associated benign hepatocellular nodules, hepatic nodules that did not fully fit the aforementioned criteria, including lesions assigned as confidence level 1 or 2 during review for lesion detection, were considered to be DN [20]. Although these criteria were presented as rough reference standards to the observers, the final decision was made by the subjective judgment of each observer. To support the results of qualitative analysis, quantitative measurements were performed by one radiologist who did not attend the qualitative analysis by measuring signal intensity (SI) of the liver, HCCs, and ICCs using operator-defined regions of interest (ROIs) in the arterial phase, portal phase, 3 min-late phase, and 10 min- and 20 min HBP. This observer was blinded to the final diagnosis of tumors. For the measurements of SIliver, each ROI area was drawn to avoid the inclusion of confounding structures such as blood vessels or artifacts. Subsequently, liver-to-lesion signal ratio (SR) was calculated as [(SIlesion − SIliver)/SIliver]. In addition, to support target appearance of ICC in qualitative analysis, we measured both signal ratio of central enhancing portion and peripheral rim of ICC on 10 min HBP.
SPSS version 18.0; SPSS, Chicago, IL, USA). An alternative-free response receiver operating characteristic (ROC) curve analysis was performed on a lesion-by-lesion basis for detection and classification of hepatic lesions [27]. The diagnostic accuracy was determined by measuring the area under the ROC curve (Az). A non-parametric analysis of clustered ROC curve data was performed, considering the possible influence of lesion clustering on diagnostic accuracy, using the method proposed by Obuchowski [28]. The sensitivities for tumor detection and classification as HCC or not HCC were calculated according to the number of lesions assigned a confidence level of 3 or 4. The specificity for HCC diagnosis was also calculated. In addition, the number of tumors that were correctly characterized according to the reference standard was calculated for each observer. The κ statistic for multiple observers was calculated to assess interobserver agreement for lesion detection and characterization. A κ value b 0.20 indicated positive but poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; and greater than 0.81, excellent agreement [29].
2.4. Statistical analysis
3.2. Detection of focal liver lesions
Statistical analyses were performed using statistical software (MedCalc version 11.4, MedCalc Software; Mariakerke, Belgium;
The calculated Az value and sensitivities for detecting all lesions are described in Table 1. In pooled data of three observers, Az value and
3. Results 3.1. Histopathologic results Of a total of 144 patients, 123 patients had 148 HCCs (size range: 0.4–10.0 cm; mean: 3.3 ± 2.6 cm), 12 hemangiomas (size range: 0.4–1.5 cm), and one LGDN (1.5 cm). Based on pathological analysis, 100 patients had one solitary HCC each, 21 patients had two HCCs each, and two patients had three HCCs each. Of 148 HCCs, seven lesions were grade I, 128 lesions were grade II, and the remaining 13 lesions were grade III according to Edmondson’s classification of HCC [30]. Nine HCCs were proven to be scirrhous HCC because they had fibrous stroma in ≥ 50% of the tumor area by histological morphometry in the greatest dimension of the cut surface [31]. The remaining 21 patients had 13 ICCs (size range: 1.8–11.0 cm; mean: 4.3 ± 2.5 cm), three HCAs (5.2 cm, 8.0 cm, and 10.5 cm), two FNHs (1.2 and 3.3 cm), one HGDN (0.7 cm), one large RN (0.7 cm), one RLH (1.4 cm), and one BDA (1.0 cm). One patient had both single HGDN and one large RN, and the remaining 20 patients had one lesion each.
Table 2 Values obtained with gadoxetic acid-enhanced MRI with diffusion weighted imaging for the lesion classification as HCC or non-HCC.
Az value Sensitivity Specificity
Observer 1
Observer 2
Observer 3
Pooled data
0.952 ± 0.018 / 95% CI: 0.912–0.978 93.9% (139) / 95% CI: 88.4–97.0 80.0% [7] / 95% CI: 62.5–90.9
0.906 ± 0.027 / 95% CI: 0.856–0.943 91.9% (136) / 95% CI: 86.0–95.5 74.3% [9] / 95% CI: 56.4–86.9
0.910 ± 0.027 / 95% CI: 0.860–0.946 91.9% (136) / 95% CI: 86.0–95.5 68.6% [11] / 95% CI: 50.6–82.6
0.922 ± 0.014 / 95% CI:0.897–0.943 92.6% (411) / 95% CI: 89.6–94.8 74.3 % [27] / 95% CI: 64.7–82.1
Note: Data represent Az values ± 1SD; numbers in parentheses and brackets represent the number of true-positive lesions and false-positive lesions, respectively.
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1207
Table 3 Results of tumor characterization with gadoxetic acid-enhanced imaging with diffusion-weighted imaging by three observers.
Observer 1 Observer 2 Observer 3
Correct diagnosis Misdiagnosis Correct diagnosis Misdiagnosis Correct diagnosis Misdiagnosis
HCC (N = 148)
ICC (N = 13)
Hemangioma (N = 12)
HCA (N = 3)
FNH (N = 2)
RLH (N = 1)
BDA (N = 1)
DN or RN (N = 3)
139 DN (5) ICC (1) Missed (3) 136 DN (9) ICC (1) Missed (2) 136 DN (5) ICC (2) Missed (5)
12 HCC (1) 11 HCC (2) 10 HCC (3)
11 HCC (1) 11 Missed (1) 12 0
2 HCC (1) 1 HCC (2) 0 HCC (3)
1 HCC (1) 0 HCC (2) 0 HCC (2)
0 HCC (1) 0 HCC (1) 0 HCC (1)
0 HCC (1) 0 HCC (1) 0 HCC (1)
2 HCC (1) 1 HCC (1) Missed (1) 1 HCC (1) Missed (1)
Note: Data represent the number of tumors. HCC, hepatocellular carcinoma; ICC, intrahepatic cholangiocarcinoma; HCA, hepatocellular adenoma; FNH, focal nodular hyperplasia; RLH, reactive lymphoid hyperplasia; BDA, bile duct adenoma; DN, dysplastic nodule; RN, regenerative nodule.
sensitivity are 0.978 ± 0.006 and 97.5%, respectively (Table 1). The lesions (size range: 0.2–0.7 cm) that were not verified on MRI was three HCCs in observer 1; two HCCs, one large RN, and one hemangioma in observer 2; and six HCCs and one large RN in observer 3. All HCCs not verified by reviewers were daughter nodules or coexisting tiny HCCs of the main tumor. On reviewing DWI, 144 HCCs
except four and all 13 ICCs showed hyperintensity. Of solid benign lesions, DWI hyperintensity was observed in 3 HCAs, 1 FNH, 1 RLH, 1 BDA, and 1 DN. There were two HCCs (0.4 cm and 0.6 cm) in two patients that were not verified by any observer (confidence rating 0). On review, one was not discernible even on retrospective review, and one was seen as faintly hypointense on HBP.
Fig. 1. Surgically confirmed 1.5-cm-sized hepatocellular carcinoma in a 61-year-old man. Axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) 3 min delay, and (D) 20 min HBP after administration of gadoxetic acid show a small hepatic mass with arterial hypervascularization (arrow) and delayed hypointensity (arrows). This tumor is clearly seen as a hyperintense and dark area (D) on single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2 and (E) on the ADC map (arrows), respectively. (F) Gross specimen confirmed a small hepatocellular carcinoma.
1208
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Fig. 2. Surgically confirmed 1.8-cm-sized intrahepatic cholangiocarcinoma in a 65-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) 3 min delay, and (C) 10 min HBP image after administration of gadoxetic acid, the tumor shows arterial hypervascularization (arrow) and delayed hypointensity (arrows). The faint hyperintense area in the center is suspicious on the 10 min HBP phase (C) but is not definitive. On single-shot echo-planar diffusion weighed imaging (D) at b = 800 s/mm2, the tumor shows clear hyperintensity with a faint hypointense area in the center (arrow). The tumor was misdiagnosed as HCC by all observers. (E) Gross specimen confirmed a small cholangiocarcinoma.
3.3. Characterization of focal liver lesions The calculated Az value and sensitivities for classifying tumors as HCC or not HCC are described in Table 2. Az values were 0.952 ± 0.018 for observer 1, 0.906 ± 0.027 for observer 2, and 0.910 ± 0.027 for observer 3, with 0.922 ± 0.014 for pooled data. Among 183 focal liver lesions, 91.3% (167/183, observer 1), 87.4% (160/183, observer 2), and 86.9% (159/183, observer 3) were correctly characterized according to the reference standard, and 89.7% of pooled data were correctly characterized (Table 3). On reviewing 148 HCCs, 140 HCCs showed arterial hyperenhancement, hypointensity on HBP, and hyperintensity on DWI (Fig. 1) [6–11]. On 3 min late phase, 137 of these 140 HCCs showed a variable degree of hypointensity, and the remaining three were isointense to liver parenchyma. Other than two to five tiny HCCs missed by each observer, five or nine HCCs (0.4–1.3 cm in diameter) were misclassified as non-HCC (DN) by each observer. On retrospective review, most of them did not clearly show arterial hyperenhance-
ment or hypointensity on late phase or HBP or hyperintensity on DWI. One scirrhous HCC was interpreted as ICC by all observers. Thus, observer 1 correctly characterized 93.9% (n = 139) of HCCs, and observers 2 and 3 correctly characterized 91.9% (n = 136) of HCCs, respectively. Specificities for HCC diagnosis were 80.0% for observer 1, 74.3% for observer 2, and 68.6% for observer 3, with a 74.3% specificity for pooled data. Of 13 ICCs, 12, 11, and 10 were correctly characterized by each observer, respectively, and between one and three (1.8–2.8 cm in diameter) were misinterpreted as HCC by each observer (Fig. 2). On reviewing 3 ICCs misinterpreted, they were observed as arterially hyperenhancing mass and showed faint central enhancement with hypointense rim on 10 min HBP and/or hyperintensity with central hypointense area on b-800 DWI. Six ICCs were considered to be hypervascular tumors because they showed arterial hyperenhancement in more than 50% of the total tumor area (Fig. 2). On 3 min late phase, two tumors were seen as mostly isointense to the liver, and four were seen as hypointense with variable degrees of central enhancing
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1209
Fig. 3. Surgically confirmed 2.0-cm-sized intrahepatic cholangiocarcinoma in a 57-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase, (B) portal venous phase, and (C) 10 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows arterial hypervascularization and hypointensity on B and C. On 10-min HBP, the tumor (arrow) shows a target appearance consisting of central enhancement and a peripheral hypointense rim. (D) On single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2, the tumor also shows a target appearance revealing peripheral hyperintensity compared to the central portion (arrow).
area. Twelve ICCs showed a centrally enhanced area and peripheral hypointense rim (target appearance) on 10 min HBP as well as a target appearance consisting of a central hypointense area and a peripheral hyperintense rim on b-800 DWI (Fig. 3). The difference in enhancement features between HCC and ICC could be supported by the results of quantitative analysis (Table 4) as HCC showed higher mean values of SR on arterial phase and portal phase, and thereafter lower SR than ICC. On 10 min HBP, the central area of ICC tended to show higher SR than tumor periphery (mean value; −0.222 ± 0.181 vs −0.341 ± 0.176), indicating target appearing enhancement. With regard to the characterization of the three HCAs and two FNHs, observer 1 misinterpreted one HCA and one FNH as HCCs. Observers 2 and 3 misinterpreted two FNHs (Fig. 5) and two or three HCAs as being HCCs. These misinterpreted lesions showed arterial hyperenhancement and hypointensity on HBP and/or signal decrease on the opposed phase compared to an in-phase chemical shift imaging indicating intralesional fat (Fig. 4). All FNHs had no central scar. One (1.2 cm) fitted HCC imaging criteria, but showed hyperintense rim on HBP. Another FNH (3.0 cm) showed no arterial hyperenhancement and hypointensity on HBP with isointensity on T1- and T2WI. RLH, BDA, and one HGDN were all misinterpreted as HCCs by all three
observers because these lesions showed arterial hyperenhancement, hypointensity on HBP, and hyperintensity on DWI. 3.4. Inter-observer agreement The kappa values were 0.713 (observer 1 vs. 2), 0.622 (observer 1 vs. 3), and 0.692 (observer 2 vs. 3) for tumor detection and 0.743 (observer 1 vs. 2), 0.743 (observer 1 vs. 3), and 0.749 (observer 2 vs. 3) for tumor characterization, thus indicating good inter-observer agreement. 4. Discussion This study was conducted in a viral hepatitis endemic population with a high risk of developing HCC. Owing to early hepatocyte uptake of gadoxetic acid, the majority of focal liver lesions, excluding lesions that preserve hepatocyte function, tend to show hypointensity on HBP and even on 3 min late phase, which might increase the chance of misdiagnosing a hypervascular lesion as HCC in a viral hepatitis endemic population. Apart from these considerations, the majority of benign tumors included were surgically resected, which possibly explains their atypical imaging features similar to those of
Table 4 Results of signal ratio of hepatocellular carcinoma and intrahepatic cholangiocarcinoma on each phase.
HCC ICC
Arterial phase
Portal phase
3 min late phase
10 min HBP
20 min HBP
0.588 ± 0.371 0.081 ± 0.299
−0.077 ± 0.098 −0.097 ± 0.169
−0.217 ± 0.141 −0.113 ± 0.139
−0.298 ± 0.196 −0.258 ± 0.170
−0.464 ± 0.074 −0.353 ± 0.151
Note: Data represent mean values ± SD. HCC, hepatocellular carcinoma; ICC, intrahepatic cholangiocarcinoma; HBP, hepatobiliary phase.
1210
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
Fig. 4. Surgically confirmed 5.2-cm-sized hepatocellular adenoma in a 33-year-old man. On axial 3D T1-weighted GRE images obtained at (A) arterial phase and (B) 20 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows arterial hypervascularization and hypointensity, respectively. (C) Opposed-phase spoiled gradient-echo MRI shows signal drop compared to the (D) in-phase image, indicating diffuse fatty infiltration of the tumor (arrow). The tumor was misdiagnosed as HCC by all observers. (E) Gross specimen confirmed a hepatocellular adenoma with fat.
hepatic malignancies. Thus, biases could have been introduced during image review or population selection. Given that lesion characterization is made not only by enhancement pattern, but also by ancillary findings on each MR sequence in daily practice, we hypothesized that a combined reading of whole MR sequences would offset such limitations and bias. With regard to tumor detection, given that the gadoxetic acid MRI protocol yielded high sensitivity (97.5% of pooled data) and accuracy (0.978 of pooled data) for detecting 183 hepatic tumors, our study reaffirmed previous studies showing promising sensitivity of the combination of gadoxetic acid MRI and DWI for detecting HCC [19,20]. This might be attributable to a mixed contribution of HBP and DWI, which offer excellent lesion-to-liver contrast. Other than two to five tiny HCCs missed by all observers, all five or nine HCCs misclassified as DN by the individual observers were daughter nodules or satellite nodules with non-clear signal change on arterial phase or HBP and DWI.
Since liver transplantation is an established treatment option for HCC, but not for ICC, precise differentiation between ICC and HCC is of paramount importance. A major concern regarding use of gadoxetic acid in HCC workup is misdiagnosis of ICC as HCC due to pseudowashout phenomenon on 3 min late phase [15]. This is particularly true for small ICC due to its tendency to demonstrate arterial hypervascularity [12,14,15]. On 3 min late phase image, four of six hypervascular ICCs were hypointense and two were isointense, whereas three HCCs were isointense. Given that small HCCs frequently show no contrast washout on portal venous or 3 min late phase images after administration of an ECF agent, as ICC does [6–9], the differentiation between HCC and ICC does not highly rely on whether tumoral enhancement is sustained on delayed phase. In practice, even when using an ECF agent, there may be uncertainty in distinguishing the two tumors. In this study, of 13 ICCs, between one and three small hypervascular ICCs (1.8–2.8 cm in diameter) were misdiagnosed as HCC by each observer. On review, three ICCs showed faint central
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
1211
Fig. 5. Surgically confirmed 1.2-cm-sized focal nodular hyperplasia in a 63-year-old woman. On axial 3D T1-weighted GRE images obtained at (A) arterial phase and (B) 20 min HBP image after administration of gadoxetic acid, the tumor (arrows) shows hyperenhancement and hypointensity, respectively. However, the tumor shows a hyperintense rim at 20 min HBP. (C) On single-shot echo-planar diffusion weighed imaging at b = 800 s/mm2, the tumor is clearly seen as hyperintense. The tumor was misdiagnosed as HCC by two observers. (D) Gross specimen confirmed a foal nodular hyperplasia.
enhancement with a hypointense rim on 10 min HBP and/or hyperintensity with central hypointense area on b-800 DWI, both of which are considered indicators of tumor with abundant fibrosis such as ICC [14,15,26]. Although the majority of ICCs were seen as mostly hypointense on 3 min late phase and/or HBP, similar to HCCs, all ICCs except one demonstrated a target appearance on 10 min HBP and b-800 DWI, suggesting ICC with fibrous stroma [14,15,26]. This was supported by the results of quantitative analysis. Thus, it is necessary for radiologists to discern the aforementioned features for accurate characterization of ICC. In addition, most ICCs tended to show stronger enhancement in the tumor periphery than in the center, which might be responsible for the correct characterization of seven scirrhous HCCs except two as HCC by all observers, although similar features in HBP and DWI might be observed in both ICC and scirrhous HCC due to their fibrotic components. For the diagnosis of FNH and HCA with standard MRI, presence of central scar or capsule, intralesional fat, tumor contour, signal intensity on T1- and T2WI as well as dynamic enhancement pattern were all considered [21–24]. However, based on findings on standard MRI, 40% of FNHs and HCAs were inconclusive because of the lack of their typical features [22]. In such case, HBP is reportedly useful to differentiate between FNH and HCA because most FNHs appear at least partially hyperintense to the liver, while most HCAs are seen as hypointense on HBP [21,22]. However, two FNHs were seen as primarily hypointense on HBP and were misdiagnosed as HCC. On review, one (1.2 cm) showed a hyperintense rim on HBP, which might be an important sign indicating FNH (Fig. 5) [21,32]. Another FNH (3.0 cm) showed neither arterial hyperenhancement nor enhancement on HBP, but displayed a serrated outer contour, which could be a differentiating feature from encapsulated HCC or HCA. Since HCA might show arterial hyperenhancement, HBP hypointensity, and a fat component, misdiagnosis of HCA as HCC is possible, as was shown in our cases. However, despite of large tumor size, HCA did not show a mosaic appearance with intratumoral septum, which is frequently seen in advanced HCC
[33,34]. Based on the above considerations, discerning ancillary features of a tumor on MRI, which is not afforded by other imaging modalities, makes accurate tumor characterization possible. RLH and BDA showed similar features to HCC, revealing the need for further large-scale evaluations to determine specific imaging features. The imaging differentiation between HCC and DN is also challenging. One HGDN was misdiagnosed as HCC by all observers. Even on review, it was compatible with hypervascular HCC, indicating incomplete differentiation between HCC and DN with the application of any imaging modality. Our study had limitations. First, it included only patients with chronic liver disease who underwent surgery, which might have caused a selection bias. In particular, the proportion of hypervascular ICCs was higher than previous studies [14,15,26]. This may be due to tumor surveillance in patients with chronic liver disease, which could lead to early detection of small ICCs that are more frequently seen as hypervascular mass than larger ICCs. However, our study did not primarily focus on tumor incidence or proportion of specific tumor type. Benign solid tumors included were all surgically resected, which may have caused an underestimation of the ability of MRI to characterize tumors. Second, issues regarding the differentiation between HCC and DN were not fully addressed because only two DNs and one large RN were macroscopically detectable. In addition, it is not possible to recruit all cirrhosisassociated benign hepatocellular nodules that were only pathologically proven but not verified on imaging and vice versa. Third, intraindividual comparison between current MRI protocol and MRI with conventional ECF agent was not made, creating uncertainty regarding the relative efficacy of the current MRI protocol for tumor characterization in chronic liver disease. In addition, a similar study using a similar MR imager and ECF MR contrast agent has not been conducted. Fourth, the result might not reflect the performance of only gadoxetic acid-enhanced MRI because reviewers were allowed to read whole MR sequences to approximate real practice.
1212
S. Kwon et al. / Magnetic Resonance Imaging 32 (2014) 1214–1222
In conclusion, despite concerns regarding limitations in tumor characterization with gadoxetic acid-enhanced MRI due to its early hepatocyte uptake, gadoxetic acid-enhanced MRI including DWI showed a reasonable performance for tumor characterization by offering high sensitivity for tumor detection in patients with chronic hepatitis or cirrhosis. It is necessary for radiologists to discern ancillary features beyond the enhancement characteristics of variable hepatic tumors provided by MRI in order to better characterize hepatic tumors.
References [1] Yuen MF, Cheng CC, Lauder IJ, Lam SK, Ooi CG, Lai CL. Early detection of hepatocellular carcinoma increases the chance of treatment: Hong Kong experience. Hepatology 2000;31:330–5. [2] Bruix J, Llovet JM. Major achievements in hepatocellular carcinoma. Lancet 2009; 373:614–6. [3] Haradome H, Grazioli L, Tinti R, et al. Additional value of gadoxetic acid-DTPAenhanced hepatobiliary phase MR imaging in the diagnosis of early-stage hepatocellular carcinoma: comparison with dynamic triple-phase multidetector CT imaging. J Magn Reson Imaging 2011;34:69–78. [4] Jung G, Breuer J, Poll LW, et al. Imaging characteristics of hepatocellular carcinoma using the hepatobiliary contrast agent GD-EOB-DTPA. Acta Radiol 2006;47:15–23. [5] Brismar TB, Dahlstrom N, Edsborg N, et al. Liver vessel enhancement by Gd-BOPTA and Gd-EOB-DTPA: a comparison in healthy volunteers. Acta Radiol 2009;50:709–15. [6] Park G, Kim YK, Kim CS, et al. Diagnostic efficacy of gadoxetic acid-enhanced MRI in the detection of hepatocellular carcinomas: comparison with gadopentetate dimeglumine. Br J Radiol 2010;83:1010–6. [7] Sun HY, Lee JM, Shin CI, et al. Gadoxetic acid-enhanced magnetic resonance imaging for differentiating small hepatocellular carcinomas (b=2 cm in diameter) from arterial enhancing pseudolesions: special emphasis on hepatobiliary phase imaging. Invest Radiol 2010;45:96–103. [8] Kim YK, Kim CS, Han YM, et al. Detection of small hepatocellular carcinoma: can gadoxetic acid-enhanced magnetic resonance imaging replace combining gadopentetate dimeglumine-enhanced and superparamagnetic iron oxideenhanced magnetic resonance imaging? Invest Radiol 2010;45:740–6. [9] Ahn SS, Kim MJ, Lim JS, et al. Added value of gadoxetic acid-enhanced hepatobiliary phase MR imaging in the diagnosis of hepatocellular carcinoma. Radiology 2010;255:459–66. [10] Kim YK, Kim CS, Han YM, et al. Detection of hepatocellular carcinoma: gadoxetic acid-enhanced 3-dimensional magnetic resonance imaging versus multidetector row computed tomography. J Comput Assist Tomogr 2009;33:844–50. [11] Kim SH, Kim SH, Lee J, et al. Gadoxetic acid-enhanced MRI versus triple-phase MDCT for the preoperative detection of hepatocellular carcinoma. Am J Roentgenol 2009;192:1675–81. [12] Kim SA, Lee JM, Lee KB, et al. Intrahepatic mass-forming cholangiocarcinomas: enhancement patterns at multiphasic CT, with special emphasis on arterial enhancement pattern-correlation with clinicopathologic findings. Radiology 2011;260:148–57. [13] Kim SJ, Lee JM, Han JK, Kim KH, Lee JY, Choi BI. Peripheral mass-forming cholangiocarcinoma in cirrhotic liver. AJR Am J Roentgenol 2007;189:1428–34. [14] Park HJ, Kim YK, Park MJ, Lee WJ. Small intrahepatic mass-forming cholangiocarcinoma: target sign on diffusion-weighted imaging for differentiation from hepatocellular carcinoma. Abdom Imaging 2013;38:793–801.
[15] Jeong HT, Kim MJ, Chung YE, Choi JY, Park YN, Kim KW. Gadoxetate disodiumenhanced MRI of mass-forming intrahepatic cholangiocarcinomas: imaginghistologic correlation. AJR Am J Roentgenol 2013;201:W603–11. [16] Kitao A, Zen Y, Matsui O, et al. Hepatocellular carcinoma: signal intensity at gadoxetic acid-enhanced MR imaging-correlation with molecular transporters and histopathologic features. Radiology 2010;256:817–26. [17] Taouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254:47–66. [18] Nasu K, Kuroki Y, Nawano S, et al. Hepatic metastases: diffusion-weighted sensitivity-encoding versus SPIO-enhanced MR imaging. Radiology 2006;239: 122–30. [19] Park MJ, Kim YK, Lee MW, et al. Small hepatocellular carcinomas: improved sensitivity by combining gadoxetic acid-enhanced and diffusion-weighted MR imaging patterns. Radiology 2012;265:104–14. [20] Park MJ, Kim YK, Lee MH, Lee JH. Validation of diagnostic criteria using gadoxetic acid-enhanced and diffusion-weighted MR imaging for small hepatocelllular carcinoma (b=2.0 cm) in patients with hepatitis-induced liver cirrhosis. Acta Radiol 2013;54:127–36. [21] Grieser C, Steffen IG, Seehofer D, et al. Histopathologically confirmed focal nodular hyperplasia of the liver: gadoxetic acid-enhanced MRI characteristics. Magn Reson Imaging 2013;31:755–60. [22] Bieze M, van den Esschert JW, Nio CY, et al. Diagnostic accuracy of MRI in differentiating hepatocellular adenoma from focal nodular hyperplasia: prospective study of the additional value of gadoxetate disodium. AJR Am J Roentgenol 2012;199:26–34. [23] Ronot M, Vilgrain V. Imaging of benign hepatocellular lesions: current concepts and recent updates. Clin Res Hepatol Gastroenterol 2014 [in press]. [24] Laumonier H, Bioulac-Sage P, Laurent C, Zucman-Rossi J, Balabaud C, Trillaud H. Hepatocellular adenomas: magnetic resonance imaging features as a function of molecular pathological classification. Hepatology 2008;48:808–18. [25] Taouli B. Diffusion-weighted MR, imaging for liver lesion characterization: a critical look. Radiology 2012;262:378–80. [26] Kim MY, Kim YK, Park HJ, Park MJ, Lee WJ, Choi D. Diagnosis of focal liver lesions with gadoxetic acid-enhanced MRI: is a shortened delay time possible by adding diffusion-weighted imaging? J Magn Reson Imaging 2014;39:31–41. [27] Chakraborty DP, Winter LH. Free-response methodology: alternate analysis and a new observer-performance experiment. Radiology 1990;174:873–81. [28] Obuchowski NA. Nonparametric analysis of clustered ROC curve data. Biometrics 1997;53:567–78. [29] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–74. [30] Edmondson HA, Steiner PE. Primary carcinoma of liver: study of 100 cases among 48,900 necropsies. Cancer 1954;7:462–503. [31] Matsuura S, Aishima S, Taguchi K, et al. 'Scirrhous' type hepatocellular carcinomas: a special reference to expression of cytokeratin 7 and hepatocyte paraffin 1. Histopathology 2005;47:382–90. [32] Fujiwara H, Sekine S, Onaya H, Shimada K, Mikata R, Arai Y. Ring-like enhancement of focal nodular hyperplasia with hepatobiliary-phase Gd-EOBDTPA-enhanced magnetic resonance imaging: radiological–pathological correlation. Jpn J Radiol 2011;29:739–43. [33] Stevens WR, Gulino SP, Batts KP, Stephens DH, Johnson CD. Mosaic pattern of hepatocellular carcinoma: histologic basis for a characteristic CT appearance. J Comput Assist Tomogr 1996;20:337–42. [34] Willatt JM, Hussain HK, Adusumilli S, Marrero JA. MR imaging of hepatocellular carcinoma in the cirrhotic liver: challenges and controversies. Radiology 2008; 247:311–30.