ª Springer Science+Business Media New York 2014
Abdominal Imaging
Abdom Imaging (2014) DOI: 10.1007/s00261-014-0244-4
Differentiation of lipid poor angiomyolipoma from hepatocellular carcinoma on gadoxetic acid-enhanced liver MR imaging Rihyeon Kim,1 Jeong Min Lee,1,2 Ijin Joo,1 Dong Ho Lee,1 Sungmin Woo,1 Joon Koo Han,1,2 Byung Ihn Choi1,2 1 2
Department of Radiology, Seoul National University Hospital, 101 Daehak-ro, Jongno-Gu, Seoul 110-744, Korea Institute of Radiation Medicine, Seoul National University Hospital, 101 Daehak-ro, Jongno-Gu, Seoul 110-744, Korea
Abstract Purpose: To investigate magnetic resonance (MR) findings of angiomyolipoma (AML) on gadoxetic acidenhanced MR imaging, and to identify features that differentiate AML from hepatocellular carcinoma (HCC) in patients with a low risk of HCC development. Methods: This retrospective study was institutional review board approved, and the requirement for informed consent was waived. Twelve patients with hepatic AML who underwent gadoxetic acid-enhanced MRI with no risk factors for HCC development were recruited. Twenty-seven patients with HCC under the same inclusion criteria were recruited as control. Two radiologists analyzed the images in consensus for morphologic features, enhancement patterns, and hepatobiliary phase (HBP) findings. All results were analyzed using the Mann–Whitney test, two-tailed Fisher exact test, and chi-square test. Results: Patients with AML were younger than those with HCC (48.8 ± 15 years for AML vs. 62.7 ± 14.2 years for HCC, p = 0.008) with female predominance, while most HCC patients were male (75% (9/12) vs. 15% (4/27), p < 0.001). The most prevalent enhancement pattern was arterial enhancement followed by hypointensity at portal or transitional phases for both AMLs (58% (7/12)) and HCCs (74% (20/27)) (p = 0.455). However, during the HBP, AMLs frequently showed more homogeneous hypointensity than HCCs (83% (10/12) vs. 41% (11/27), p = 0.018). When compared with the signal intensity of the spleen, the mean relative signal intensity of the AML was 91.2 ± 15.4%, while in HCCs, it was 128.7 ± 40% (p < 0.001).
Correspondence to: Jeong Min Lee; email: [email protected]
Conclusions: Although AMLs showed similar enhancement patterns to HCCs during the dynamic phases of gadoxetic acid-enhanced MRI, using characteristic MR features of AML during the HBP and demographic differences, one can better differentiate AML from HCC. Key words: Hepatic angiomyolipoma—Hepatocellular carcinoma—Gadoxetic acid-enhanced MR— Hepatobiliary phase
Angiomyolipoma (AML) is a hypervascular hepatic tumor which typically develops in middle-aged women [1]. The most important radiographic features of AML include the presence of blood vessels and mature adipose tissue [2], however, the presence of fat may not be specific as there are other types of liver tumors which also contain fat, such as lipoma, myelolipoma, hepatocellular carcinoma (HCC), and metastasis [3]. Moreover, the epitheloid type of AML has been shown to contain no or a minimal amount of macroscopic fat [2]. Thus, until now, a correct preoperative diagnosis of AML was only able to be demonstrated in 20–52% of cases, in which the most commonly made misdiagnosis was HCC, as the enhancement patterns of AMLs have often been shown to mimic HCCs [4–9]. Given that 80–90% of HCCs develop in a population at a high risk for HCC development, such as those with viral hepatitis B- or C-induced cirrhosis or alcoholic cirrhosis, the high-risk clinical setting may be an important clue for the diagnosis of HCC [10, 11]. However, the remaining 10–20% of HCCs could also develop in a low-risk population as well as in those with nonalcoholic liver disease. Furthermore, as hypervascularity and the presence of internal fat components are common in both AMLs and HCCs, differentiation between these
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
two entities in patients without liver cirrhosis has remained difficult [12–15]. Considering the different treatment strategies employed for these two diseases, with conservative treatment generally recommended for hepatic AMLs owing to its benign disease course and aggressive treatment recommended for HCCs as they are malignant tumors with a much poorer prognosis [16, 17], differential diagnosis of these two entities, particularly in the clinical setting involving a low-risk factor for HCC, is of crucial importance. Recently, gadoxetic acid (Gd-EOB-DTPA; Primovist, Bayer Healthcare, Berlin, Germany) has gained wide acceptance as a contrast medium for liver magnetic resonance (MR) imaging, as hepatobiliary phase (HBP) images provide additional valuable information for tumor characterization [18–20]. Until now, however, there have been only sporadic reports describing the imaging features of AML on hepatocyte-specific contrast media-enhanced MR imaging [21, 22], and comparison of imaging features between AMLs and HCCs on gadoxetic acid-enhanced MR imaging in a population with a low risk for HCC development has not been well described in the literature. Therefore, the purpose of this study was to investigate the MR findings of AML on gadoxetic acid-enhanced MR imaging and to identify imaging features that can differentiate AML from HCC in patients with a low risk for HCC development.
Materials and methods This retrospective study was approved by our institutional review board, and the requirement for informed consent was waived.
Fig. 1.
Patients Detailed patient selection and inclusion and exclusion criteria are described in Fig. 1. A thorough search of the electronic medical record system of our hospital using a search term of AML resulted in 23 cases of pathologically diagnosed AMLs from November 2004 to November 2013. The inclusion criteria were as follows: (1) absence of a risk factor for HCC development, such as HBV or HCV infection and alcoholic liver cirrhosis, (2) absence of a history of surgery or interventional treatment before the MR examination, and (3) presence of gadoxetic acid MR taken within 3 months prior to the pathologic diagnosis. Finally, a total of 12 patients (mean age, 48.8 years; range 24–70 years), including 3 men (mean age, 56.7 years; range 32–70 years) and 9 women (mean age, 46.2 years; range 24–60 years), diagnosed between 2008 and 2013 were included in this study. The pathologic diagnosis of AML was confirmed through surgery (n = 6) or image-guided percutaneous biopsy (n = 6). To establish a control group, we searched the electronic medical record system of our hospital for an initial diagnosis of nodular-type HCC from November 2012 to November 2013 and found 279 cases. After applying the same inclusion criteria, a total of 27 patients (mean age, 62.7 years; range 32–85 years), including 23 men (mean age, 62 years; range 32–77 years) and 4 women (mean age, 67.3 years; range 33–85 years), were selected also with the absence of risk factors for HCC development. The pathologic diagnosis of HCC was made through surgery (n = 19) or image-guided percutaneous biopsy (n = 8).
Flow diagram of patient selection, inclusion and exclusion criteria.
Scanner 1, 1.5T, SignaHDx or HDxt; Scanner 2, 3.0T, Verio or Ingenia; TR, Repetition time; TE, echo time; BH, breath-hold; FSE, fast spin echo; RT, Respiratory-triggered; FS, fat suppression; GRE, gradient-recalled echo; 2D, two dimensional; 3D, three dimensional; DW, diffusion-weighted; SSSE, single-shot sine-echo; EPI, echo planar imaging
11 180 15 180 0 0 50 0 3 7 4.8 7 384 9 307 136 9 136 3.4/1.2 5000/52 4.2/1.9 5000/52
320 9 256 136 9 136
90 90 70 10 10 0 20 20 20 7 7 3 7 7 7 384 9 307 448 9 269 320 9 285 384 9 192 512 9 384 512 9 192 982/163 3000/91 4/1.3, 2.3
Scanner 1 Scanner 2 Scanner 1 Scanner 2 Scanner 1 Scanner 1 Scanner 2 Scanner 1
Scanner 2
Flip angle (°) Intersection gap (%) Section thickness (mm) Matrix (mm 9 mm) TR (ms)/TE (ms)
P/86.9 3000/97.1 140/2.4, 5.8
BH T2 single-shot FSE RT FS-T2 FSE BH 2D or 3D T1 In/ Opposed GRE FS 3D T1-weighted GRE DW SSSE EPI
All MR images were retrospectively reviewed on a picture archiving and communication system workstation (Marosis; Marotech, Seoul, Korea). The reviewers were blinded to the pathologic diagnosis of each lesion, however, information regarding tumor location was provided so as not to be confused with other pathologic lesions.
Sequence
Image analysis
Table 1. Parameters for gadoxetic acid-enhanced liver MRI
Gadoxetic acid (PrimovistÒ)-enhanced MR imaging was performed prior to the pathologic diagnosis [mean time before diagnosis, 18 days (range 1–87 days)]. Since the study was retrospectively designed, a variety of MR imaging machines were used. A total of 29 MRI exams were performed at our hospital using 1.5T units (n = 18: SignaHDx or HDxt, GE Medical Systems, Milwaukee, Wis) and 3T units (n = 11: Ingenia [Philips Healthcare, Best, the Netherlands] or Verio [Siemens healthcare, Erlangen, Germany]). The remaining 10 MRI images were acquired at outside hospital; 4 cases from 1.5T units [SignaHDx (GE Medical Systems, Milwaukee, Wis) or Intera (Philips Healthcare, Best, the Netherlands)] and the other 6 cases from 3.0T units [Ingenia (Philips Healthcare, Best, the Netherlands) or Verio (Siemens healthcare, Erlangen, Germany) or Skyra or Triotm (Siemens healthcare, Erlangen, Germany)]. All MR images were obtained in the axial plane using either an 8-channel torso phased-array coil at the 1.5T unit or a 12- or 32-channel coil at the 3.0T unit, centered on the liver. A rectangular field of view of 320 to 350 mm was adjusted for each patient’s body size and was held constant for all sequences. The baseline MR imaging sequences were as follows: a respiratory-triggered T2weighted fast spin-echo sequence, a breath-hold heavily T2-weighted half-Fourier acquisition sequence, diffusion-weighted images (DWIs), a breath-hold T1-weighted dual-echo gradient-recalled echo (GRE) sequence (In/opposed-phase), and a breath-hold T1-weighted fatsuppressed (FS) GRE sequence. The parameters of each pulse sequence are summarized in Table 1. Dynamic imaging was performed by bolus tracking method with a standard dose of gadoxetic acid (0.025 mmol/kg, Primovist; Bayer-Schering, Berlin, Germany) at a rate of 1.5 mL/s, immediately followed by a 30-mL saline flush using a power injector (Spectris SolarisÒ EP, MEDRAD Inc., Warrendale, PA, USA). Dynamic imaging was carried out using the same FS 3D GRE sequence with the arterial phase (AP) scanned 7 seconds after the contrast arrived at the distal thoracic aorta, and subsequent portal phase (PP), transitional phase (TP), and HBP scans conducted 60 seconds, 3 minutes, and 20 minutes after beginning contrast medium administration, respectively.
Scanner 2
Image acquisition
140 140 9
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
Images of a 48-year-old woman with pathologically confirmed angiomyolipoma in the left lobe of the liver. (A) Axial T2-weighted image shows a 38 mm-sized roundshaped nodular lesion with hyperintensity compared to the surrounding liver parenchyma at segment 2. When comparing opposed-phase (C) to in-phase images (B), no definite intratumoral microscopic fat component is suspected. (D) At fatsuppressed T1-weighted image, this lesion shows hypointensity compared to the liver parenchyma. (E) At gadoxetic acid-enhanced T1-weighted image during the arterial phase (AP), this lesion shows strong enhancement. During the portal phase (PP) (F) and subsequent transitional phase (TP) (G), the lesion becomes hypointense in signal compared to the liver parenchyma, which was interpreted as the ‘‘wash in and wash out’’ enhancement pattern. (H) At 20 min-delayed hepatobiliary phase (HBP), this lesion shows marked hypointensity compared to the liver parenchyma as well as the spleen with a clear margin. According to the quantitative analysis, the relative signal intensity of the tumor compared to the spleen (SR T/S = SItumor/SIspleen) was 90.0%, and the enhancement ratio [ER = (SIHBP - SIprecontrast)/SIprecontrast] was 31.5% in this case.
b Fig. 2.
portal vein [27]), and (h) tumor necrosis (high T1 SI and low SI with absence of enhancement). The enhancement degree of the lesions compared to the adjacent hepatic parenchyma at each dynamic phase was documented. On AP, the presence of punctate or curved dilated feeding arteries was evaluated [2]. In addition, the presence of early draining veins and conspicuous vessels originating from the tumor draining to the portal vein or hepatic vein was assessed [15]. Temporal enhancement was categorized as wash-in and wash-out (strong enhancement at AP followed by hypoenhancement at PP or TP [28]), persistent enhancement (contrast enhancement remaining invariable throughout the AP, PP, and TP [29]), and the others. The margin of the tumor at the HBP was analyzed as well- or ill-defined. Finally, the relative SI of the tumor compared to the spleen was documented as being of hypo-, iso-, or hyperintensity at the HBP.
Quantitative analysis Qualitative analysis Qualitative image analysis was made by a consensus of two board-certified abdominal radiologists (with 8 and 9 years of experience, respectively). The diameter of each tumor was measured at the axial plane on T2- weighted images, and the multiplicity of the lesion was documented. The margin was evaluated as either sharp or illdefined and the contour as oval or round, lobulating or irregular. In addition, the homogeneity of the signal intensity (SI) was determined. The relative signal intensity of the tumor compared to the liver parenchyma was categorized as hypo-, iso-, or hyperintensity at T2weighted image. If the tumor was heterogeneous, the predominant (>50%) signal intensity was documented. The presence of a microscopic intratumoral fat portion manifested as diffuse signal drop and the india ink artifact at the mass–liver interface or within a hepatic mass suggesting that the macroscopic fat component was assessed at T1-weighted dual-echo image. The diffusion restriction was determined from a high b-value (b > 800) diffusion-weighted image and apparent diffusion coefficient map [23]. In addition, the following ancillary findings were documented: (a) peripheral capsule (a discernible peripheral rim structure surrounding at least half of the tumor [24]); (b) mosaic appearance (areas of variable SI, created by a confluence of small nodules separated by thin septa and necrosis [25]); (c) satellite nodules; (d) LN enlargement (larger than 1 cm at axial image); (e) hemorrhagic foci (focal high SI at precontrast FS T1 image [26]); (f) transient hepatic intensity difference surrounding the tumor (patchy wedge-shaped area of enhancement involving a hepatic subsegment [25]); (g) portal vein thrombus (intraluminal structures inside the
Quantitative analysis using region-of-interest (ROI) measurements was performed by a third radiologist (with 2 years of experience in abdominal imaging) independent of the qualitative analysis. In order to measure the ratio of T2 SI of the tumor compared to the liver, one radiologist measured SI of the tumors by drawing a manually defined oval-shaped ROI to cover the tumor as large as possible, and drew another ROI identical in size to that used in the measurement of the hepatic tumor at the adjacent liver parenchyma trying to avoid the vessels or other artifacts. Then, T2 SI ratio of tumor-to-liver was calculated as the ratio of tumor SI and the liver parenchyma on T2-weighted images. Additional measurement of tumor SI was performed at both the precontrast T1 image and the HBP. Thereafter, another oval-shaped ROI of approximately 170 mm2 (range 150–244 mm2) was placed on the liver parenchyma as well as the spleen, taking care to avoid vessels or other areas of focal signal changes at the HBP. After measurement of the four signal intensities, three parameters were calculated: tumor-to-liver SI ratio (SRT/L), i.e., SR = SItumor/SIliver; tumor-to-spleen SI ratio (SR T/S), i.e., SR = SItumor/SIspleen; and enhancement ratio of the tumor (ER), where ER = (SIHBP - SIprecontrast)/ SIprecontrast [30].
Statistical analysis Differences in size of the tumor, age of each group, and the quantitative MR features between AML and HCC were analyzed using the Mann–Whitney test. Other nominal variants were compared using the Chi-square test and Fisher exact test, as appropriate. A two-sided
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
significance level of 5% was considered to indicate a statistically significant difference for all analyses. Statistical analyses were performed using the SPSS 19.0 software package (SPSS, Chicago, Ill).
Results Patient demographics Mean age of patients in the AML group was lower than that of patients in the HCC group (48.8 ± 15 years vs. 62.7 ± 14.2 years, p = 0.008). Female predominance was observed in the AML group, while most HCC patients were male (75% (9/12) vs. 15% (4/27), p < 0.001).
Morphologic features of AMLs and HCCs on unenhanced MR imaging Mean diameter of the largest lesion was 3.8 ± 2.1 cm in the AML group and 5.5 ± 3.2 cm in the HCC group (p = 0.080). AMLs (33%, 4/12) showed a mosaic appearance less frequently than HCCs (74%, 20/27; p = 0.031), and more commonly showed a homogeneous T2 SI than HCCs: 75% (9/12) vs. 37% (10/27) (p = 0.029). Findings of the multiplicity of lesions, margin, contour, peripheral capsule, satellite nodules, LN enlargement, necrosis, hemorrhagic foci, the presence of both microscopic and macroscopic fat presence at dual-echo image, diffusion restriction, transient hepatic intensity difference, feeding
artery dilatation, multiple aneurysmal artery, early draining veins, portal vein involvement, and the relative T2 SI ratio of the tumor than liver did not differ between the two groups (p > 0.05) (Figs. 2A–D, 3A–D). Detailed morphologic features of both tumors are summarized in Table 2.
Enhancing features of AMLs and HCCs at dynamic phases All AMLs (100%, 12/12) and 85% (23/27) of HCCs showed strong enhancement during the AP (p = 0.292). In addition, both AMLs and HCCs demonstrated iso- to hypointensity at the PP compared to the liver parenchyma [92% (11/12) vs. 100% (27/27), p = 0.258] as well as at the TP [100% (12/12) vs. 96% (26/27), p = 0.794]. Therefore, the most prevalent enhancement pattern of both AMLs and HCCs was hyperenhancement at the AP (wash-in) and hypoenhancement at the PP or TP (wash-out) (58% (7/12) in AMLs and 74% (20/27) in HCCs, p = 0.455) (Figs. 2D, E, F, G,3D, E, F, G). Results of enhancement analysis are summarized in Tables 2 and 3.
Image features of AMLs and HCCs at the hepatobiliary phase Qualitative analysis on hepatobiliary phase. All AMLs (100%, 12/12) and 81% (22/27) of HCCs showed hypointensity on the HBP compared to the liver parenchyma
Table 2. Morphologic characteristics and enhancing features of AMLs and HCCs Qualitative variables
AML (n = 12)
HCC (n = 27)
p value
Mean size ± SD (cm) Lesion multiplicity Sharp margin Oval or round Contour Peripheral capsule Mosaic appearance Satellite nodules LN enlargement Hemorrhagic foci Tumor necrosis Microscopic fat presence at dual echoa India ink artifact at dual echo Diffusion restrictionb T2 SI ratio of the tumor than liver Homogeneity of T2 SI THID Dilated feeding artery Multiple aneurysmal arteries Early draining vein Portal vein thrombus Temporal enhancement pattern Wash-in wash-out Persistent enhancement
3.8 ± 2.1 2 (17%) 11 (92%) 11 (92%) 5 (42%) 4 (33%) 1 (8%) 0 (0%) 1 (8%) 1 (8%) 7 (58%) 3 (25%) 12 (100%) 3.47 ± 3.0 9 (75%) 1 (8%) 6 (50%) 0 (0%) 3 (25%) 0 (0%)
5.5 ± 3.2 4 (15%) 17 (63%) 14 (52%) 19 (70%) 20 (74%) 5 (19%) 0 (0%) 8 (30%) 10 (37%) 9 (33%) 2 (7%) 26 (96%) 2.48 ± 1.1 10 (37%) 7 (26%) 18 (67%) 1(4%) 2 (7%) 1 (4%)
0.080 1.000 0.122 0.055 0.153 0.031 0.645 1.000 0.228 0.064 0.174 0.159 1.000 0.285 0.029 0.394 0.478 1.000 0.159 1.000 0.455
7 (58%) 5 (42%)
20 (74%) 3 (11%)
p values from Mann–Whitney test for size of the tumor and T2 SI ratio of the tumor than liver, p values from Chi-square test or Fisher exact test for other variables a Presence of fat was defined when the lesion showed a signal drop on opposed-phase images compared with in-phase images b Diffusion restriction was determined on ADC maps AML, angiomyolipoma; HCC, hepatocellular carcinoma; SD, standard deviation; LN, lymph node; SI, signal intensity; THID, transient hepatic intensity difference
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
Table 3. Relative signal intensity and enhancement degrees of AMLs and HCCs Groups
AML (n = 12) HCC (n = 27)
Signal intensitya
Hyperintensity Isointensity Hypointensity Hyperintensity Isointensity Hypointensity p value
Precontrast imaging
Postcontrast imaging
T2WI
T1WI
AP
PP
TP
HBP
12 (100%) 0 (0%) 0 (0%) 25 (93%) 2 (7%) 0 (0%) 1.000
1 (8%) 1 (8%) 10 (83%) 0 (0%) 5 (19%) 22 (81%) 0.245
12 (100%) 0 (0%) 0 (0%) 23 (85%) 4 (15%) 0 (0%) 0.292
1 (8%) 4 (33%) 7 (58%) 0 (0%) 7 (26%) 20 (74%) 0.258
0 (0%) 1 (8%) 11 (92%) 1 (4%) 2 (7%) 24 (89%) 0.794
0 (0%) 0 (0%) 12 (100%) 2 (8%) 3 (11%) 22 (81%) 0.280
p values from Chi-square test or Fisher exact test a Signal intensity refers to relative signal intensity of tumors compared to the liver parenchyma AML, angiomyolipoma; HCC, hepatocellular carcinoma; AP, arterial phase; PP, portal phase; TP, transitional phase; HBP, hepatobiliary phase
Table 4. Hepatobiliary phase findings of gadoxetic acid-enhanced MR imaging for AMLs and HCCs Variables
AML (n = 12)
HCC (n = 27)
p value
Homogeneity of HBP SI Well-defined tumor margin HBP SI compared to spleen Hypo or Iso SI Hyper SI Quantitative analysis results SRT/L (%)a SRT/S (%)a ER (%)a
10 (83%) 11 (92%)
11 (41%) 18 (67%)
0.018 0.131
Abdominal Imaging
Abdom Imaging (2014) DOI: 10.1007/s00261-014-0244-4
Differentiation of lipid poor angiomyolipoma from hepatocellular carcinoma on gadoxetic acid-enhanced liver MR imaging Rihyeon Kim,1 Jeong Min Lee,1,2 Ijin Joo,1 Dong Ho Lee,1 Sungmin Woo,1 Joon Koo Han,1,2 Byung Ihn Choi1,2 1 2
Department of Radiology, Seoul National University Hospital, 101 Daehak-ro, Jongno-Gu, Seoul 110-744, Korea Institute of Radiation Medicine, Seoul National University Hospital, 101 Daehak-ro, Jongno-Gu, Seoul 110-744, Korea
Abstract Purpose: To investigate magnetic resonance (MR) findings of angiomyolipoma (AML) on gadoxetic acidenhanced MR imaging, and to identify features that differentiate AML from hepatocellular carcinoma (HCC) in patients with a low risk of HCC development. Methods: This retrospective study was institutional review board approved, and the requirement for informed consent was waived. Twelve patients with hepatic AML who underwent gadoxetic acid-enhanced MRI with no risk factors for HCC development were recruited. Twenty-seven patients with HCC under the same inclusion criteria were recruited as control. Two radiologists analyzed the images in consensus for morphologic features, enhancement patterns, and hepatobiliary phase (HBP) findings. All results were analyzed using the Mann–Whitney test, two-tailed Fisher exact test, and chi-square test. Results: Patients with AML were younger than those with HCC (48.8 ± 15 years for AML vs. 62.7 ± 14.2 years for HCC, p = 0.008) with female predominance, while most HCC patients were male (75% (9/12) vs. 15% (4/27), p < 0.001). The most prevalent enhancement pattern was arterial enhancement followed by hypointensity at portal or transitional phases for both AMLs (58% (7/12)) and HCCs (74% (20/27)) (p = 0.455). However, during the HBP, AMLs frequently showed more homogeneous hypointensity than HCCs (83% (10/12) vs. 41% (11/27), p = 0.018). When compared with the signal intensity of the spleen, the mean relative signal intensity of the AML was 91.2 ± 15.4%, while in HCCs, it was 128.7 ± 40% (p < 0.001).
Correspondence to: Jeong Min Lee; email: [email protected]
Conclusions: Although AMLs showed similar enhancement patterns to HCCs during the dynamic phases of gadoxetic acid-enhanced MRI, using characteristic MR features of AML during the HBP and demographic differences, one can better differentiate AML from HCC. Key words: Hepatic angiomyolipoma—Hepatocellular carcinoma—Gadoxetic acid-enhanced MR— Hepatobiliary phase
Angiomyolipoma (AML) is a hypervascular hepatic tumor which typically develops in middle-aged women [1]. The most important radiographic features of AML include the presence of blood vessels and mature adipose tissue [2], however, the presence of fat may not be specific as there are other types of liver tumors which also contain fat, such as lipoma, myelolipoma, hepatocellular carcinoma (HCC), and metastasis [3]. Moreover, the epitheloid type of AML has been shown to contain no or a minimal amount of macroscopic fat [2]. Thus, until now, a correct preoperative diagnosis of AML was only able to be demonstrated in 20–52% of cases, in which the most commonly made misdiagnosis was HCC, as the enhancement patterns of AMLs have often been shown to mimic HCCs [4–9]. Given that 80–90% of HCCs develop in a population at a high risk for HCC development, such as those with viral hepatitis B- or C-induced cirrhosis or alcoholic cirrhosis, the high-risk clinical setting may be an important clue for the diagnosis of HCC [10, 11]. However, the remaining 10–20% of HCCs could also develop in a low-risk population as well as in those with nonalcoholic liver disease. Furthermore, as hypervascularity and the presence of internal fat components are common in both AMLs and HCCs, differentiation between these
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
two entities in patients without liver cirrhosis has remained difficult [12–15]. Considering the different treatment strategies employed for these two diseases, with conservative treatment generally recommended for hepatic AMLs owing to its benign disease course and aggressive treatment recommended for HCCs as they are malignant tumors with a much poorer prognosis [16, 17], differential diagnosis of these two entities, particularly in the clinical setting involving a low-risk factor for HCC, is of crucial importance. Recently, gadoxetic acid (Gd-EOB-DTPA; Primovist, Bayer Healthcare, Berlin, Germany) has gained wide acceptance as a contrast medium for liver magnetic resonance (MR) imaging, as hepatobiliary phase (HBP) images provide additional valuable information for tumor characterization [18–20]. Until now, however, there have been only sporadic reports describing the imaging features of AML on hepatocyte-specific contrast media-enhanced MR imaging [21, 22], and comparison of imaging features between AMLs and HCCs on gadoxetic acid-enhanced MR imaging in a population with a low risk for HCC development has not been well described in the literature. Therefore, the purpose of this study was to investigate the MR findings of AML on gadoxetic acid-enhanced MR imaging and to identify imaging features that can differentiate AML from HCC in patients with a low risk for HCC development.
Materials and methods This retrospective study was approved by our institutional review board, and the requirement for informed consent was waived.
Fig. 1.
Patients Detailed patient selection and inclusion and exclusion criteria are described in Fig. 1. A thorough search of the electronic medical record system of our hospital using a search term of AML resulted in 23 cases of pathologically diagnosed AMLs from November 2004 to November 2013. The inclusion criteria were as follows: (1) absence of a risk factor for HCC development, such as HBV or HCV infection and alcoholic liver cirrhosis, (2) absence of a history of surgery or interventional treatment before the MR examination, and (3) presence of gadoxetic acid MR taken within 3 months prior to the pathologic diagnosis. Finally, a total of 12 patients (mean age, 48.8 years; range 24–70 years), including 3 men (mean age, 56.7 years; range 32–70 years) and 9 women (mean age, 46.2 years; range 24–60 years), diagnosed between 2008 and 2013 were included in this study. The pathologic diagnosis of AML was confirmed through surgery (n = 6) or image-guided percutaneous biopsy (n = 6). To establish a control group, we searched the electronic medical record system of our hospital for an initial diagnosis of nodular-type HCC from November 2012 to November 2013 and found 279 cases. After applying the same inclusion criteria, a total of 27 patients (mean age, 62.7 years; range 32–85 years), including 23 men (mean age, 62 years; range 32–77 years) and 4 women (mean age, 67.3 years; range 33–85 years), were selected also with the absence of risk factors for HCC development. The pathologic diagnosis of HCC was made through surgery (n = 19) or image-guided percutaneous biopsy (n = 8).
Flow diagram of patient selection, inclusion and exclusion criteria.
Scanner 1, 1.5T, SignaHDx or HDxt; Scanner 2, 3.0T, Verio or Ingenia; TR, Repetition time; TE, echo time; BH, breath-hold; FSE, fast spin echo; RT, Respiratory-triggered; FS, fat suppression; GRE, gradient-recalled echo; 2D, two dimensional; 3D, three dimensional; DW, diffusion-weighted; SSSE, single-shot sine-echo; EPI, echo planar imaging
11 180 15 180 0 0 50 0 3 7 4.8 7 384 9 307 136 9 136 3.4/1.2 5000/52 4.2/1.9 5000/52
320 9 256 136 9 136
90 90 70 10 10 0 20 20 20 7 7 3 7 7 7 384 9 307 448 9 269 320 9 285 384 9 192 512 9 384 512 9 192 982/163 3000/91 4/1.3, 2.3
Scanner 1 Scanner 2 Scanner 1 Scanner 2 Scanner 1 Scanner 1 Scanner 2 Scanner 1
Scanner 2
Flip angle (°) Intersection gap (%) Section thickness (mm) Matrix (mm 9 mm) TR (ms)/TE (ms)
P/86.9 3000/97.1 140/2.4, 5.8
BH T2 single-shot FSE RT FS-T2 FSE BH 2D or 3D T1 In/ Opposed GRE FS 3D T1-weighted GRE DW SSSE EPI
All MR images were retrospectively reviewed on a picture archiving and communication system workstation (Marosis; Marotech, Seoul, Korea). The reviewers were blinded to the pathologic diagnosis of each lesion, however, information regarding tumor location was provided so as not to be confused with other pathologic lesions.
Sequence
Image analysis
Table 1. Parameters for gadoxetic acid-enhanced liver MRI
Gadoxetic acid (PrimovistÒ)-enhanced MR imaging was performed prior to the pathologic diagnosis [mean time before diagnosis, 18 days (range 1–87 days)]. Since the study was retrospectively designed, a variety of MR imaging machines were used. A total of 29 MRI exams were performed at our hospital using 1.5T units (n = 18: SignaHDx or HDxt, GE Medical Systems, Milwaukee, Wis) and 3T units (n = 11: Ingenia [Philips Healthcare, Best, the Netherlands] or Verio [Siemens healthcare, Erlangen, Germany]). The remaining 10 MRI images were acquired at outside hospital; 4 cases from 1.5T units [SignaHDx (GE Medical Systems, Milwaukee, Wis) or Intera (Philips Healthcare, Best, the Netherlands)] and the other 6 cases from 3.0T units [Ingenia (Philips Healthcare, Best, the Netherlands) or Verio (Siemens healthcare, Erlangen, Germany) or Skyra or Triotm (Siemens healthcare, Erlangen, Germany)]. All MR images were obtained in the axial plane using either an 8-channel torso phased-array coil at the 1.5T unit or a 12- or 32-channel coil at the 3.0T unit, centered on the liver. A rectangular field of view of 320 to 350 mm was adjusted for each patient’s body size and was held constant for all sequences. The baseline MR imaging sequences were as follows: a respiratory-triggered T2weighted fast spin-echo sequence, a breath-hold heavily T2-weighted half-Fourier acquisition sequence, diffusion-weighted images (DWIs), a breath-hold T1-weighted dual-echo gradient-recalled echo (GRE) sequence (In/opposed-phase), and a breath-hold T1-weighted fatsuppressed (FS) GRE sequence. The parameters of each pulse sequence are summarized in Table 1. Dynamic imaging was performed by bolus tracking method with a standard dose of gadoxetic acid (0.025 mmol/kg, Primovist; Bayer-Schering, Berlin, Germany) at a rate of 1.5 mL/s, immediately followed by a 30-mL saline flush using a power injector (Spectris SolarisÒ EP, MEDRAD Inc., Warrendale, PA, USA). Dynamic imaging was carried out using the same FS 3D GRE sequence with the arterial phase (AP) scanned 7 seconds after the contrast arrived at the distal thoracic aorta, and subsequent portal phase (PP), transitional phase (TP), and HBP scans conducted 60 seconds, 3 minutes, and 20 minutes after beginning contrast medium administration, respectively.
Scanner 2
Image acquisition
140 140 9
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
Images of a 48-year-old woman with pathologically confirmed angiomyolipoma in the left lobe of the liver. (A) Axial T2-weighted image shows a 38 mm-sized roundshaped nodular lesion with hyperintensity compared to the surrounding liver parenchyma at segment 2. When comparing opposed-phase (C) to in-phase images (B), no definite intratumoral microscopic fat component is suspected. (D) At fatsuppressed T1-weighted image, this lesion shows hypointensity compared to the liver parenchyma. (E) At gadoxetic acid-enhanced T1-weighted image during the arterial phase (AP), this lesion shows strong enhancement. During the portal phase (PP) (F) and subsequent transitional phase (TP) (G), the lesion becomes hypointense in signal compared to the liver parenchyma, which was interpreted as the ‘‘wash in and wash out’’ enhancement pattern. (H) At 20 min-delayed hepatobiliary phase (HBP), this lesion shows marked hypointensity compared to the liver parenchyma as well as the spleen with a clear margin. According to the quantitative analysis, the relative signal intensity of the tumor compared to the spleen (SR T/S = SItumor/SIspleen) was 90.0%, and the enhancement ratio [ER = (SIHBP - SIprecontrast)/SIprecontrast] was 31.5% in this case.
b Fig. 2.
portal vein [27]), and (h) tumor necrosis (high T1 SI and low SI with absence of enhancement). The enhancement degree of the lesions compared to the adjacent hepatic parenchyma at each dynamic phase was documented. On AP, the presence of punctate or curved dilated feeding arteries was evaluated [2]. In addition, the presence of early draining veins and conspicuous vessels originating from the tumor draining to the portal vein or hepatic vein was assessed [15]. Temporal enhancement was categorized as wash-in and wash-out (strong enhancement at AP followed by hypoenhancement at PP or TP [28]), persistent enhancement (contrast enhancement remaining invariable throughout the AP, PP, and TP [29]), and the others. The margin of the tumor at the HBP was analyzed as well- or ill-defined. Finally, the relative SI of the tumor compared to the spleen was documented as being of hypo-, iso-, or hyperintensity at the HBP.
Quantitative analysis Qualitative analysis Qualitative image analysis was made by a consensus of two board-certified abdominal radiologists (with 8 and 9 years of experience, respectively). The diameter of each tumor was measured at the axial plane on T2- weighted images, and the multiplicity of the lesion was documented. The margin was evaluated as either sharp or illdefined and the contour as oval or round, lobulating or irregular. In addition, the homogeneity of the signal intensity (SI) was determined. The relative signal intensity of the tumor compared to the liver parenchyma was categorized as hypo-, iso-, or hyperintensity at T2weighted image. If the tumor was heterogeneous, the predominant (>50%) signal intensity was documented. The presence of a microscopic intratumoral fat portion manifested as diffuse signal drop and the india ink artifact at the mass–liver interface or within a hepatic mass suggesting that the macroscopic fat component was assessed at T1-weighted dual-echo image. The diffusion restriction was determined from a high b-value (b > 800) diffusion-weighted image and apparent diffusion coefficient map [23]. In addition, the following ancillary findings were documented: (a) peripheral capsule (a discernible peripheral rim structure surrounding at least half of the tumor [24]); (b) mosaic appearance (areas of variable SI, created by a confluence of small nodules separated by thin septa and necrosis [25]); (c) satellite nodules; (d) LN enlargement (larger than 1 cm at axial image); (e) hemorrhagic foci (focal high SI at precontrast FS T1 image [26]); (f) transient hepatic intensity difference surrounding the tumor (patchy wedge-shaped area of enhancement involving a hepatic subsegment [25]); (g) portal vein thrombus (intraluminal structures inside the
Quantitative analysis using region-of-interest (ROI) measurements was performed by a third radiologist (with 2 years of experience in abdominal imaging) independent of the qualitative analysis. In order to measure the ratio of T2 SI of the tumor compared to the liver, one radiologist measured SI of the tumors by drawing a manually defined oval-shaped ROI to cover the tumor as large as possible, and drew another ROI identical in size to that used in the measurement of the hepatic tumor at the adjacent liver parenchyma trying to avoid the vessels or other artifacts. Then, T2 SI ratio of tumor-to-liver was calculated as the ratio of tumor SI and the liver parenchyma on T2-weighted images. Additional measurement of tumor SI was performed at both the precontrast T1 image and the HBP. Thereafter, another oval-shaped ROI of approximately 170 mm2 (range 150–244 mm2) was placed on the liver parenchyma as well as the spleen, taking care to avoid vessels or other areas of focal signal changes at the HBP. After measurement of the four signal intensities, three parameters were calculated: tumor-to-liver SI ratio (SRT/L), i.e., SR = SItumor/SIliver; tumor-to-spleen SI ratio (SR T/S), i.e., SR = SItumor/SIspleen; and enhancement ratio of the tumor (ER), where ER = (SIHBP - SIprecontrast)/ SIprecontrast [30].
Statistical analysis Differences in size of the tumor, age of each group, and the quantitative MR features between AML and HCC were analyzed using the Mann–Whitney test. Other nominal variants were compared using the Chi-square test and Fisher exact test, as appropriate. A two-sided
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
significance level of 5% was considered to indicate a statistically significant difference for all analyses. Statistical analyses were performed using the SPSS 19.0 software package (SPSS, Chicago, Ill).
Results Patient demographics Mean age of patients in the AML group was lower than that of patients in the HCC group (48.8 ± 15 years vs. 62.7 ± 14.2 years, p = 0.008). Female predominance was observed in the AML group, while most HCC patients were male (75% (9/12) vs. 15% (4/27), p < 0.001).
Morphologic features of AMLs and HCCs on unenhanced MR imaging Mean diameter of the largest lesion was 3.8 ± 2.1 cm in the AML group and 5.5 ± 3.2 cm in the HCC group (p = 0.080). AMLs (33%, 4/12) showed a mosaic appearance less frequently than HCCs (74%, 20/27; p = 0.031), and more commonly showed a homogeneous T2 SI than HCCs: 75% (9/12) vs. 37% (10/27) (p = 0.029). Findings of the multiplicity of lesions, margin, contour, peripheral capsule, satellite nodules, LN enlargement, necrosis, hemorrhagic foci, the presence of both microscopic and macroscopic fat presence at dual-echo image, diffusion restriction, transient hepatic intensity difference, feeding
artery dilatation, multiple aneurysmal artery, early draining veins, portal vein involvement, and the relative T2 SI ratio of the tumor than liver did not differ between the two groups (p > 0.05) (Figs. 2A–D, 3A–D). Detailed morphologic features of both tumors are summarized in Table 2.
Enhancing features of AMLs and HCCs at dynamic phases All AMLs (100%, 12/12) and 85% (23/27) of HCCs showed strong enhancement during the AP (p = 0.292). In addition, both AMLs and HCCs demonstrated iso- to hypointensity at the PP compared to the liver parenchyma [92% (11/12) vs. 100% (27/27), p = 0.258] as well as at the TP [100% (12/12) vs. 96% (26/27), p = 0.794]. Therefore, the most prevalent enhancement pattern of both AMLs and HCCs was hyperenhancement at the AP (wash-in) and hypoenhancement at the PP or TP (wash-out) (58% (7/12) in AMLs and 74% (20/27) in HCCs, p = 0.455) (Figs. 2D, E, F, G,3D, E, F, G). Results of enhancement analysis are summarized in Tables 2 and 3.
Image features of AMLs and HCCs at the hepatobiliary phase Qualitative analysis on hepatobiliary phase. All AMLs (100%, 12/12) and 81% (22/27) of HCCs showed hypointensity on the HBP compared to the liver parenchyma
Table 2. Morphologic characteristics and enhancing features of AMLs and HCCs Qualitative variables
AML (n = 12)
HCC (n = 27)
p value
Mean size ± SD (cm) Lesion multiplicity Sharp margin Oval or round Contour Peripheral capsule Mosaic appearance Satellite nodules LN enlargement Hemorrhagic foci Tumor necrosis Microscopic fat presence at dual echoa India ink artifact at dual echo Diffusion restrictionb T2 SI ratio of the tumor than liver Homogeneity of T2 SI THID Dilated feeding artery Multiple aneurysmal arteries Early draining vein Portal vein thrombus Temporal enhancement pattern Wash-in wash-out Persistent enhancement
3.8 ± 2.1 2 (17%) 11 (92%) 11 (92%) 5 (42%) 4 (33%) 1 (8%) 0 (0%) 1 (8%) 1 (8%) 7 (58%) 3 (25%) 12 (100%) 3.47 ± 3.0 9 (75%) 1 (8%) 6 (50%) 0 (0%) 3 (25%) 0 (0%)
5.5 ± 3.2 4 (15%) 17 (63%) 14 (52%) 19 (70%) 20 (74%) 5 (19%) 0 (0%) 8 (30%) 10 (37%) 9 (33%) 2 (7%) 26 (96%) 2.48 ± 1.1 10 (37%) 7 (26%) 18 (67%) 1(4%) 2 (7%) 1 (4%)
0.080 1.000 0.122 0.055 0.153 0.031 0.645 1.000 0.228 0.064 0.174 0.159 1.000 0.285 0.029 0.394 0.478 1.000 0.159 1.000 0.455
7 (58%) 5 (42%)
20 (74%) 3 (11%)
p values from Mann–Whitney test for size of the tumor and T2 SI ratio of the tumor than liver, p values from Chi-square test or Fisher exact test for other variables a Presence of fat was defined when the lesion showed a signal drop on opposed-phase images compared with in-phase images b Diffusion restriction was determined on ADC maps AML, angiomyolipoma; HCC, hepatocellular carcinoma; SD, standard deviation; LN, lymph node; SI, signal intensity; THID, transient hepatic intensity difference
R. Kim et al.: Differentiation of lipid poor angiomyolipoma
Table 3. Relative signal intensity and enhancement degrees of AMLs and HCCs Groups
AML (n = 12) HCC (n = 27)
Signal intensitya
Hyperintensity Isointensity Hypointensity Hyperintensity Isointensity Hypointensity p value
Precontrast imaging
Postcontrast imaging
T2WI
T1WI
AP
PP
TP
HBP
12 (100%) 0 (0%) 0 (0%) 25 (93%) 2 (7%) 0 (0%) 1.000
1 (8%) 1 (8%) 10 (83%) 0 (0%) 5 (19%) 22 (81%) 0.245
12 (100%) 0 (0%) 0 (0%) 23 (85%) 4 (15%) 0 (0%) 0.292
1 (8%) 4 (33%) 7 (58%) 0 (0%) 7 (26%) 20 (74%) 0.258
0 (0%) 1 (8%) 11 (92%) 1 (4%) 2 (7%) 24 (89%) 0.794
0 (0%) 0 (0%) 12 (100%) 2 (8%) 3 (11%) 22 (81%) 0.280
p values from Chi-square test or Fisher exact test a Signal intensity refers to relative signal intensity of tumors compared to the liver parenchyma AML, angiomyolipoma; HCC, hepatocellular carcinoma; AP, arterial phase; PP, portal phase; TP, transitional phase; HBP, hepatobiliary phase
Table 4. Hepatobiliary phase findings of gadoxetic acid-enhanced MR imaging for AMLs and HCCs Variables
AML (n = 12)
HCC (n = 27)
p value
Homogeneity of HBP SI Well-defined tumor margin HBP SI compared to spleen Hypo or Iso SI Hyper SI Quantitative analysis results SRT/L (%)a SRT/S (%)a ER (%)a
10 (83%) 11 (92%)
11 (41%) 18 (67%)
0.018 0.131
