Contrast-enhanced magnetic resonance (MR) imaging is a highly

Transcription

1 ORIGINAL ARTICLE Accurate Identification of True Benign Lesions on Enhanced Dynamic and Hepatobiliary Phase Magnetic Resonance Imaging After Gadobenate Dimeglumine Administration Giovanni Morana, MD,* Luigi Grazioli, MD, Miles A. Kirchin, PhD, Maria Pia Bondioni, MD, Niccolò Faccioli, MD, Alessandro Guarise, MD,* and Günther Schneider, MD Purpose: To evaluate hepatobiliary phase magnetic resonance imaging with gadobenate dimeglumine for differentiation of benign hypervascular liver lesions from malignant or high-risk lesions. Methods and Materials: Retrospective assessment was performed of 550 patients with 910 hypervascular lesions (302 focal nodular hyperplasia FNH, 82 nodular regenerative hyperplasia NRH, 59 hepatic adenoma or liver adenomatosis HA/LA, 329 hepatocellular carcinomas HCC, 12 fibrolamellar-hcc FL-HCC, 21 peripheral cholangiocarcinomas PCC, 105 metastases). Imaging was performed before and during the arterial, portal-venous, equilibrium, and hepatobiliary phases after gadobenate dimeglumine administration (0.05 mmol/kg). Histologic confirmation was available for 1 lesion per patient, except for patients with suspected FNH (diagnosis based on characteristic enhancement/follow-up). Lesion differentiation (benign/malignant) on the basis of contrast washout and lesion enhancement (hypo-/iso-/hyperintensity) was assessed (sensitivity, specificity, accuracy, PPV, and NPV) relative to histology or final diagnosis. Results: On portal-venous or equilibrium phase images, washout was not seen for 208 of 526 (39.5%) malignant (HCC, FL-HCC, PCC, metastases) and high-risk (HA/LA) lesions. Conversely, only 5 of 384 (1.3%) true benign lesions (FNH/NRH) showed washout. Taking washout as indicating malignancy, the sensitivity, specificity, and accuracy for malignant lesion identification during these phases was 61.8%, 98.7%, and 77.4%. On hepatobiliary phase images, 289 of 302 FNH, 82 of 82 NRH, 1 of 59 HA or LA, 62 of 341 HCC or FL-HCC, and 2 of 105 metastases were hyperintense or isointense. Taking iso- or hyperintensity as an indication for lesion benignity, the sensitivity, specificity, accuracy, PPV, and NPV for benign lesion identification was 96.6%, 87.6%, 91.4%, 85.1%, and 97.3%, respectively. Conclusions: Hepatobiliary phase imaging with gadobenate dimeglumine is accurate for distinguishing benign lesions from malignant or high-risk lesions. Biopsy should be considered for hypointense lesions on hepatobiliary phase images after gadobenate dimeglumine. Key Words: focal liver lesions, magnetic resonance imaging, contrast agents, gadobenate dimeglumine (MultiHance), hepatobiliary phase MR imaging (Invest Radiol 2011;46: ) Received May 18, 2010; accepted for publication (after revision) September 11, From the *Department of Diagnostic Radiology, Ospedale Cà Foncello, Treviso, Italy; Department of Radiology, Spedali Civili Brescia, University of Brescia, Brescia, Italy; Worldwide Medical & Regulatory Affairs, Bracco Imaging SpA, Milan, Italy; Department of Radiology, University of Verona, Verona, Italy; and Department of Diagnostic and Interventional Radiology, University Hospital of Saarland, Homburg/Saar, Germany. Reprints: Giovanni Morana, MD, Department of Diagnostic Radiology, Ospedale Cà Foncello, Treviso, Italy. gmorana@ulss.tv.it. Copyright 2011 by Lippincott Williams & Wilkins ISSN: /11/ Contrast-enhanced magnetic resonance (MR) imaging is a highly accurate noninvasive imaging modality for the detection and characterization of solid hypervascular focal liver lesions (FLLs), and is invariably the imaging method of choice for improved differential diagnosis in cases of equivocal or indeterminate lesions on ultrasound or computed tomography (CT). 1 4 However, it is not always possible to accurately diagnose a given lesion on conventional T1-weighted (T1w) dynamic phase imaging because of overlapping enhancement patterns between different lesion types. 5 Moreover, the frequent atypical appearance of certain lesion types might further complicate the diagnosis. The development of gadolinium-based contrast agents with liver-specific properties has markedly improved the accuracy of MR imaging for the identification and characterization of FLLs Gadobenate dimeglumine (MultiHance; Bracco Imaging SpA, Milan, Italy) is a gadolinium-based MR contrast agent which has the combination of the properties of a conventional nonspecific gadolinium-based agent and that of an agent targeted specifically to hepatocytes This feature permits improved lesion detection relative to that achievable on unenhanced and dynamic phase imaging alone, 11,12 and an additional means for lesion characterization on the basis of the functional characteristics of lesions; lesions that contain functioning hepatocytes are able to take up the Gd-BOPTA contrast-effective molecule of gadobenate dimeglumine in a manner similar to that of normal liver parenchyma, whereas lesions that do not contain functioning hepatocytes are generally unable to take up Gd-BOPTA, and thus appear hypointense relative to enhanced normal liver parenchyma on delayed T1w hepatobiliary phase images Previous studies have shown that this differential hepatobiliary phase uptake is of particular benefit in the case of focal nodular hyperplasia (FNH), both for the improved characterization of lesions with atypical enhancement features on dynamic phase imaging 17 and for the accurate differentiation of FNH from hepatic adenoma (HA). 18 Unfortunately, the spectrum of hypervascular FLLs encountered in routine practice extends beyond benign FNH. 22 Among the most common hypervascular lesions, FNH and nodular regenerative hyperplasia (NRH) are considered truly benign because these lesions show no propensity for malignant transformation and are rarely associated with complications such as rupture, hemorrhage, or torsion, which typically necessitate emergency intervention. 23 Consequently, these lesions are usually managed conservatively with patients undergoing routine follow-up. Conversely, other common hypervascular lesions (eg, hepatocellular carcinoma HCC, peripheral cholangiocarcinoma PCC, liver metastases from primary extrahepatic tumors) may be highly malignant in nature and frequently prone to complications. These lesions usually require prompt surgical or ablative treatment whenever feasible and whenever such treatment would benefit the patient. A distinct type of solid hypervascular lesion Investigative Radiology Volume 46, Number 4, April

2 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 is HA. Although benign in nature, these lesions are frequently associated with rupture and intralesional hemorrhage, particularly if larger than 5 cm in size. Consequently, these lesions are considered high risk, and surgical resection is usually indicated for larger lesions The purpose of the present retrospective analysis was to determine the diagnostic accuracy of hepatobiliary phase MR imaging with gadobenate dimeglumine for the differentiation of true benign liver lesions (FNH, NRH) from malignant and high-risk hypervascular lesions in patients referred for MR imaging of the liver as part of clinical routine examination. METHODS AND MATERIALS Study Population, Lesions, Confirmation A total of 550 consecutive patients (290 women, 260 men; mean age SD, years; range, 7 86 years) referred for MR imaging of the liver at 1 of 4 centers in Europe and who had one or more solid hypervascular FLLs confirmed histologically or by means of routine follow-up were evaluated. All patients were evaluated as part of routine clinical practice, and were informed of the nature of the MR imaging examination and the reason why it was being performed. Approval for the retrospective analysis of patient TABLE 1. Patient Demographic Characteristics and Summary of Evaluated Lesions and Approach to Lesion Diagnosis Patients Lesions Basis for Lesion Diagnosis Lesion Type Number Male/Female Age Number Size (cm) Range (cm) Follow-Up Biopsy Resection TACE OLT Autopsy FNH / NRH 16 9/ HA/LA 37 (31/6) 7/ HCC/FL-HCC 225 (218/7) 173/ PCC 20 8/ Metastases 49 28/ Total 551* 260/291* *One female patient had both FNH and HA lesions. TACE indicates transarterial chemoembolization; OLT, orthotopic liver transplant; FNH, focal nodular hyperplasia; NRH, nodular regenerative hyperplasia; HA/LA, hepatic adenoma/liver adenomatosis; HCC, hepatocellular carcinomas; FL-HCC, fibrolamellar-hcc; PCC, peripheral cholangiocarcinomas. TABLE 2. Definitions for Analysis of Diagnostic Performance Imaging Phase Portal-venous and equilibrium phases: washout as a sign of malignancy Hepatobiliary phase: hyper/isointensity as a sign of benignity Enhancement Feature True Benign Lesions (FNH; NRH) Malignant/High-Risk Lesions (HCC/FL-HCC; PCC; Metastases; HA/LA) Washout FP TP Lack of washout TN FN Hyperintense TP FP Isointense TP FP Hypointense homogeneous FN TN Inhomogeneous hypo/isointense FN TN Inhomogeneous hypo/hyperintense FN TN FNH indicates focal nodular hyperplasia; NRH, nodular regenerative hyperplasia; HCC, hepatocellular carcinomas; FL-HCC, fibrolamellar- HCC; PCC, peripheral cholangiocarcinomas; HA/LA, hepatic adenoma/liver adenomatosis; FP, false positive; TN, true negative; TP, true positive, FN, false negative. TABLE 3. Lesion Appearance on Unenhanced Images and T1w Dynamic Phase Images After Gadobenate Dimeglumine Administration Precontrast T2w Precontrast T1w Lesion Type Hyperintensity Isointensity Hypointensity N.S. Hyperintensity Isointensity Hypointensity N.S. FNH (n 302) NRH (n 82) HA/LA (n 59) HCC/FL-HCC (n 341) PCC (n 21) Metastases (n 105) Total (n 910) N.S. indicates not seen; T1w, T1-weighted; T2w, T2-weighted; FNH, focal nodular hyperplasia; NRH, nodular regenerative hyperplasia; HA/LA, hepatic adenoma/liver adenomatosis; HCC, hepatocellular carcinomas; FL-HCC, fibrolamellar-hcc; PCC, peripheral cholangiocarcinomas Lippincott Williams & Wilkins

3 Investigative Radiology Volume 46, Number 4, April 2011 data was obtained from the Ethics committee at each center. Written informed consent from individual patients was not required for this retrospective analysis. MR examinations were performed consecutively at each center between March 2000 and September 2007, with patient follow-up extended to March 2010 for patients managed conservatively (ie, for patients undergoing follow-up for diagnosed FNH or NRH). All of the evaluated patients underwent unenhanced T2- weighted (T2w) and T1w imaging, followed by T1w dynamic and delayed hepatobiliary phase liver imaging after gadobenate dimeglumine administration. Patients with obvious capillary or cavernous hemangiomas were not included; these lesions are readily diagnosed on unenhanced T2w and postcontrast T1w dynamic phase images because of their characteristic enhancement patterns, 27 and thus delayed hepatobiliary phase images are not typically acquired for these lesions. Similarly, patients with hypovascular liver lesions on postcontrast T1w arterial phase images were not evaluated, as these lesions are mostly malignant and referred for biopsy if not immediate intervention. 28 Demographic characteristics of the patient population and a summary of the lesions included in the analysis and the approach to lesion diagnosis are given in Table 1. A total of 910 hypervascular lesions comprising 302 FNH, 82 NRH, 59 HA, 6 liver adenomatosis (LA), 329 HCC, 12 fibrolamellar-hcc FL-HCC, 21 PCC, and 105 hypervascular metastases were evaluated. The mean size of evaluated lesions was comparable, ranging between cm for NRH and cm for PCC (Table 1). The primary cancers in the patients with hepatic metastases were neuroendocrine pancreatic tumor (20 lesions in 6 patients), renal cell carcinoma (18 lesions in 6 patients), breast cancer (15 lesions in 7 patients), carcinoid (11 lesions in 6 patients), insulinoma (9 lesions in 2 patients), melanoma (7 lesions in 6 patients), colon cancer (7 lesions in 5 patients), lung adenocarcinoma (5 lesions in 2 patients), thyroid cancer (3 lesions in 1 patient), gastrointestinal stromal tumor (2 lesions in 2 patients), gastrinoma (2 lesions in 1 patient), ovarian cancer (2 lesions in 1 patient), esophageal cancer (1 lesion in 1 patient), Merkel cell tumor (1 lesion in 1 patient), rhinopharyngeal cancer (1 lesion in 1 patient), and non small-cell lung cancer (1 lesion in 1 patient). All evaluated lesions were confirmed histologically, unless the lesion enhancement patterns on dynamic and delayed hepatobiliary phase images in conjunction with clinical history and any other pertinent diagnostic information (ie, findings from any previous imaging studies) were strongly indicative of a benign lesion. Lesions strongly suspected to be benign on the basis of characteristic morphologic and enhancement features (eg, typical FNH and NRH) were included in the present analysis only if follow-up information for at least 2 years was available. 29 Atypical FNH (ie, lesions that demonstrated abnormal features such as heterogeneous enhancement; hypointensity in the portal-venous or equilibrium phases; absent central scar in larger lesions, typically 3 cm; scar hypointensity on T2w images; or the presence of a pseudocapsule, hemorrhage, or necrosis) or NRH were referred for biopsy or in some cases resected (Table 1). All lesions suspected to be malignant were confirmed histologically after biopsy or resection, or displayed identical behavior to previously biopsied lesions in the same patient during transarterial chemoembolization. Only histologically confirmed lesions were included in the analysis; additional lesions in the same patient that were not evaluated histologically were not included regardless of whether the enhancement behavior was identical to that of the biopsied lesion. Therefore, no more than 3 lesions per patient were evaluated, even among patients with multiple ostensibly identical lesions. For the purposes of the present analysis, LA lesions were grouped with HA lesions because LA lesions are histologically and radiologically identical to HA lesions. 30,31 Similarly, FL-HCC lesions were grouped with the HCC lesions for subsequent analysis. MR Imaging A similar MR imaging protocol was used at each institution. All patients were imaged on a superconducting imager (Magnetom Symphony or MagnetomVision, Siemens Medical Systems, Erlangen, Germany) operating at 1.5 Tesla, using a body-array coil. MR imaging was performed using T2w Turbo Spin Echo sequences with or without fat saturation (T2wTSE; TR/TE 4000 ms/ ms, flip angle FA 150 degree, echo train length 29) or a T2w half-fourier acquisition single-shot turbo spin-echo (HASTE) sequence (T2wHASTE; TR/TE /74 ms, FA 180 degree) and T1w Gradient Echo in-phase and out-of-phase sequences (T1wGRE; TR/TE ms/4.7 ms for in-phase/2.6 for out-of-phase, FA 70 degree) or a T1w volumetric interpolated breath-hold examination sequence (T1wVIBE; Repetition time (TR)/Echo time (TE) 6.2 ms/2.5 ms, FA 15 degree). Images were acquired before the administration of contrast agent (T2wTSE and T1wGRE or T1wVIBE images); during the dynamic phase of contrast agent enhancement (T1wGRE or T1wVIBE images only) at 25 to 30 seconds (arterial phase), 70 to 90 seconds (portal-venous phase), and 3 to 5 minutes (equilibrium phase) following the intravenous bolus (2 2.5 ml/s) administration of gadobenate dimeglumine at a dose of 0.05 mmol/kg bodyweight; and during a later delayed hepatobiliary phase (T1wGRE images only) between 1 and 3 hours postinjection. Images acquired in the postcontrast delayed phase were acquired with or without fat suppression. The slice thickness was 6 to 7 mm for unenhanced images and images acquired during the delayed TABLE 3. (Continued) T1w Arterial Phase T1w Portal-Venous Phase T1w Equilibrium Phase Hyperintensity Isointensity Hypointensity N.S. Hyperintensity Isointensity Hypointensity N.S. Hyperintensity Isointensity Hypointensity N.S Lippincott Williams & Wilkins 227

4 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 hepatobiliary phase, and between 2.5 and 7 mm depending on the sequence (T1wVIBE or T1wGRE) for image acquisition during the postcontrast dynamic phase. T2wHASTE images of the entire liver were acquired either as a single slab with a total breath-hold acquisition time of 19 to 21 seconds or in 2 slabs of 11 slices with an acquisition time of 14 seconds each. T2wTSE and T1wGRE images of the entire liver were acquired in single breath-hold acquisitions of 19 to 23 seconds. The T1wVIBE images were acquired with a single breath-hold acquisition of 18 seconds. A matrix size of was employed with a rectangular field-of-view of 350 to 420 mm. Image Evaluation All MR images at each center were evaluated by a senior radiologist (XX, YY, ZZ) with at least 18 years experience in liver imaging and more than 12 years specific experience of liver MRI with gadobenate dimeglumine. All detected lesions were assessed in terms of signal intensity enhancement on dynamic (arterial, portalvenous, and equilibrium phases) and delayed (1 3 hours) hepatobiliary phase images. Lesions were classified primarily as hypointense, isointense, or hyperintense relative to the surrounding normal liver parenchyma in the same imaging phase. In the portal-venous and equilibrium phases, lesions which showed a signal intensity lower than the surrounding normal parenchyma were considered to have demonstrated contrast washout. Secondary determinations were made of enhancement homogeneity (homogeneous, inhomogeneous) and of the presence or absence of features characteristic of specific lesion types (eg, central scar in the case of FNH). Such features are frequently useful for lesion characterization, especially for benign lesions, enabling conservative management whenever possible. Statistical Analysis Definitions of lesion features used for determinations of diagnostic performance are given in Table 2. Determinations of diagnostic performance using contrast washout as indicative of malignancy during the portal-venous and equilibrium phases and lesion hyper- or isointensity as indicative of lesion benignity during the hepatobiliary phase were performed in terms of sensitivity (TP/ TP FN ), specificity (TN/ TN FP ), accuracy ( TP TN / all lesions ), positive predictive value (PPV; TP/ TP FP ), and negative predictive value (NPV; TN/ TN FN ). RESULTS Precontrast Unenhanced Imaging The appearance of the 910 evaluated lesions on precontrast T2w and T1w images is summarized in Table 3. Detected lesions on unenhanced T2w images were predominantly hyperintense or isointense to the surrounding normal liver parenchyma, although most PCC lesions (18 of 21; 85.7%) were hypointense. Of the 38 lesions not seen on unenhanced T2w images, 31 (81.6%) were benign FNH or NRH. The appearance of lesions on unenhanced T1w images was more diverse, although most lesions were hypointense or isointense. Exceptions were NRH which in most cases appeared hyperintense and HCC whose appearance ranged from markedly hypointense to markedly hyperintense. A total of 47 lesions were not visible on unenhanced T1w images. In most cases, these were again benign FNH or NRH. Although unenhanced imaging frequently provided hints to the diagnosis of a given lesion, it was insufficient in itself to permit reliable differentiation of benign from malignant disease. Dynamic Phase Imaging Because only hypervascular lesions were included in this analysis, all evaluated lesions were hyperintense against the surrounding normal parenchyma on arterial (25 30 inches) phase images (Table 3). Although in most cases (687 of 910 lesions 75.5% ) the arterial phase enhancement was homogeneous, inhomogeneous hyperintensity reflecting areas of internal lesion necrosis, fibrosis, or hemorrhage was noted in 21.2% (193 of 910) of lesions overall, predominantly among HCC or FL-HCC (Fig. 1), PCC, metastases (Fig. 2), and HA or LA. Peripheral rim enhancement, which is frequently a characteristic feature of hypervascular metastases and PCC on dynamic MRI, 29,30 was noted in 3.3% (30 of 910) of lesions overall (11 of % PCC, 14 of % metastases (Fig. 3), and 5 of % HCC; Table 4). On portal-venous (70 90 inches) phase images, only 392 of 910 (43.1%) lesions were homogeneously or inhomogeneously hyperintense, whereas 283 of 910 (31.1%) lesions were isointense and 235 of 910 (25.8%) lesions were hypointense (Table 3). Comparison among individual lesion types revealed that although 40% or more of all true malignant lesions and 25.4% (15 of 59) of HA or LA were already hypointense during the portal-venous phase, between 27.0% and 42.9% of these lesions were still homogeneously or inhomogeneously hyperintense. A similarly high proportion of true malignant and high-risk lesions retained a hyperintense or isointense appearance on equilibrium phase acquisitions at 3-5 postinjection; overall, 57.6% (34 of 59) of HA or LA lesions, 31.1% (106 of 341) of HCC or FL-HCC lesions, 48.6% (51 of 105) of metastases, and 81.0% (17 of 21) of PCC lesions were homogeneously or inhomogeneously hyperintense or isointense during this phase (Table 3). Among these lesions, more HA or LA, HCC or FL-HCC, and metastases appeared isointense than hyperintense, reflecting continued contrast agent washout. Conversely, most PCC lesions were predominantly hyperintense during the equilibrium phase because of contrast agent pooling. Assuming that washout in the portal-venous and equilibrium phase images is indicative of a true malignant lesion (HCC, FL- HCC, PCC, or metastasis) or a high-risk lesion (HA or LA), the overall sensitivity, specificity, accuracy, PPV, and NPV for the differentiation of benign from malignant or high-risk lesions on portal-venous and equilibrium phase imaging was 61.8%, 98.7%, 77.4%, 98.5%, and 65.3%, respectively. Despite the predominantly hyperintense or isointense appearance of the FNH and NRH lesions throughout the dynamic phase of image acquisition, the high proportion of malignant and HA or LA lesions showing similar enhancement behavior precluded reliable differentiation of benign and malignant lesions on the basis of dynamic phase imaging alone. Delayed Hepatobiliary Phase Imaging The overall appearance of the 910 evaluated lesions during the delayed hepatobiliary phase of contrast enhancement is summarized in Table 5. Although 289 of 302 (95.7%) FNH (Figs. 4, 5) and all (82 of % ) NRH had a hyperintense or isointense appearance, all but 62 HCC or FL-HCC, 2 metastases, and 1 HA lesion appeared homogeneously or inhomogeneously hypointense (Figs. 5 7). Regarding the 2 metastases, both lesions demonstrated a high degree of internal necrosis and appeared inhomogeneously hyperintense on delayed phase images because of clear contrast agent pooling within the lesion. Conversely, the inhomogeneous isointense appearance of the solitary HA reflected the presence of extensive internal hemorrhage in a patient with severe hepatic steatosis. As regards to the 13 FNH lesions that had an overall hypointense appearance on delayed phase images, this was in each case due to fibrosis within the lesion. Assuming that lesion hyperintensity or isointensity on delayed hepatobiliary phase images is indicative of a true benign lesion (FNH or NRH) and that lesion hypointensity indicates either a true malignant lesion (HCC, FL-HCC, PCC or metastasis) or a high-risk Lippincott Williams & Wilkins

5 Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 1. A 17-year-old girl with a large, lobulated tumor in the left liver lobe. The lesion (arrow) appears inhomogeneously isointense or slightly hypointense on unenhanced T1w images (A) as well as on T2w TSE (B) and HASTE (C) images. Intense contrast enhancement during the T1w arterial phase (D) after administration of 0.05 mmol/kg gadobenate dimeglumine, with no significant washout during the portal-venous (E) and equilibrium (F) phases, is suggestive of the dynamic contrast enhancement pattern displayed by FNH. On delayed (2 hour) T1w hepatobiliary phase images (G), the lesion appears hypointense against the enhanced normal parenchyma, indicating a tumor without functioning hepatocytes that is unable to take up Gd-BOPTA. Histology following resection confirmed the lesion as a fibrolamellar carcinoma Lippincott Williams & Wilkins 229

6 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 2. A 59-year-old woman with multiple hypervascular metastases from a primary neuroendocrine tumor. The unenhanced T1w image (A) reveals 2 large hypointense lesions (arrows) in the right liver lobe. On T2w TSE (B) images, the lesions are almost isointense with the normal parenchyma. The large lesion (arrow) appears inhomogeneously slightly hyperintense during the arterial phase (C) and iso- to hyperintense during the portal-venous phase (D). During the equilibrium phase (E), the lesions appear inhomogeneously hypointense. On the delayed (1 hour) T1w hepatobiliary phase image (F), the lesions appear strongly inhomogeneously hypointense. Contrast agent pooling (arrow) is evident within the large lesion. The delayed T1w hepatobiliary phase image also reveals the presence of 2 additional lesions (arrows) in the left liver lobe. lesion (HA or LA), the overall sensitivity, specificity, accuracy, PPV, and NPV for the differentiation of benign from malignant lesions on hepatobiliary phase imaging in this analysis was 96.6%, 87.6%, 91.4%, 85.1%, and 97.3%, respectively (Table 6). Given that inhomogeneous or peripheral enhancement during the arterial phase is often a strong indicator of lesion malignancy, a similar assessment was performed for just the 687 hypervascular lesions that demonstrated homogeneous enhancement on arterial phase imaging. Similar overall results were obtained (98.7%, 85.4%, 92.7%, 89.2%, and 98.1%, for sensitivity, specificity, accuracy, PPV, and NPV, respectively; Table 6). Finally, because contrast washout during the portal-venous and equilibrium phases is often also indicative of lesion malignancy, assessment was similarly performed for just the 580 hypervascular lesions that did not demonstrate washout during these phases. Again, the results obtained were similar to those of the overall lesion population (96.8%, 78.1%, 90.3%, 89.3%, and 92.9%, for sensitivity, specificity, accuracy, PPV, and NPV, respectively; Table 6) Lippincott Williams & Wilkins

7 Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 3. A 55-year-old man with hypervascular metastases from a primary intestinal carcinoid. A small lesion (arrow) appears as homogeneously hypointense on the unenhanced T1w image (A) and as slightly hyperintense on the T2w TSE image (B). During the T1w arterial phase (C) after administration of 0.05 mmol/kg gadobenate dimeglumine, the lesion appears slightly hyperintense. The lesion appears isointense and hypointense on the portal-venous (D) and equilibrium (E) phase images, respectively. In the equilibrium phase, another small hypointense lesion (arrow) can be seen in the sixth segment. On the delayed (2 hour) T1w hepatobiliary phase image (F), the lesions are both strongly hypointense against the enhanced normal liver parenchyma. During this phase, another small hypointense lesion (arrowhead) can be seen Lippincott Williams & Wilkins 231

8 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 TABLE 4. Homogeneity of Lesions on Arterial Phase Images After Gadobenate Dimeglumine Administration No. Arterial Phase Enhancement Lesion Type Lesions Homogenous Inhomogeneous Peripheral FNH NRH HA/LA HCC/FL-HCC PCC Metastases Total FNH indicates focal nodular hyperplasia; NRH, nodular regenerative hyperplasia; HA/LA, hepatic adenoma/liver adenomatosis; HCC, hepatocellular carcinomas; FL-HCC, fibrolamellar-hcc; PCC, peripheral cholangiocarcinomas. TABLE 5. Appearance of Lesions on T1w Delayed Hepatobiliary Phase Images After Gadobenate Dimeglumine Administration Lesion Type No. Lesions Hyperintense Isointense Hypointense FNH NRH HA/LA HCC/FL-HCC PCC Metastases Total T1w indicates T1-weighted; FNH, focal nodular hyperplasia; NRH, nodular regenerative hyperplasia; HA/LA, hepatic adenoma/liver adenomatosis; HCC, hepatocellular carcinomas; FL-HCC, fibrolamellar-hcc; PCC, peripheral cholangiocarcinomas. DISCUSSION In the 12 years that gadobenate dimeglumine has been approved in Europe for MR imaging of the liver, numerous studies have highlighted the value of delayed hepatobiliary phase acquisitions not only for the improved detection of metastatic or small satellite lesions, 11,12 but also for the improved characterization of equivocal or otherwise atypical lesions on routine dynamic phase imaging Moreover, comparative studies have also shown that hepatobiliary phase imaging with gadobenate dimeglumine is superior to MR imaging with liver-specific superparamagnetic iron oxide particles for both the detection and characterization of liver lesions. 12,35,36 Recently, Kim et al have shown that delayed hepatobiliary phase acquisitions are particularly beneficial for the improved characterization of HCC in cirrhotic livers. 16 At variance with previous studies that focused on gadobenate dimeglumine-enhanced MR imaging of specific lesion types, 12,15 18,35,36 our analysis addressed the overall value of delayed hepatobiliary phase imaging with gadobenate dimeglumine for routine clinical differentiation of true benign lesions for which conservative treatment is typically indicated from malignant and high-risk lesions, for which therapeutic intervention is usually necessary whenever possible. Assuming that a hyperintense or isointense appearance on hepatobiliary phase imaging indicates a lesion with functioning hepatocytes that is able to take up the Gd-BOPTA contrast-effective molecule of gadobenate dimeglumine 17,18 in a similar manner to that of surrounding normal liver parenchyma, and that a hypointense appearance indicates a lesion without functioning hepatocytes that is unable to take up Gd-BOPTA, we determined overall values for sensitivity, specificity, accuracy, PPV, and NPV of 96.6%, 87.6%, 91.4%, 85.1%, and 97.3%, respectively. On the basis of these findings, an appropriate rule of thumb may be to consider a hypointense lesion on delayed hepatobiliary phase imaging in normal liver as suspicious for malignancy for which confirmatory histologic information is required, whereas a hyperintense or isointense lesion in the absence of other suspicious findings (eg, elevated -1 fetoprotein in case of HCC 37 ) is more likely to be benign in nature, and thus a candidate for regular follow-up. Interesting exceptions are HA and LA lesions which are benign in nature but which appear hypointense on delayed hepatobiliary phase imaging because they lack biliary ductules, 38 and thus are unable to take up and excrete Gd-BOPTA through the hepatobiliary route. However, these lesions are highly susceptible to hemorrhage and thus are frequently considered for surgical resection anyway, in much the same way as malignant lesions In agreement with previous findings, 18 all but one of the HA or LA lesions included in this study were hypointense on delayed phase images; the one exception was an inhomogeneously isointense lesion in a patient with severe hepatic steatosis. Apart from the severe steatosis in this patient, another possible explanation for the isointense appearance of the single HA in this study was that a gadobenate dimeglumine dose of just 0.05 mmol/kg bodyweight was used because this is the dose approved in Europe for MR imaging of the liver. Although the R1 relaxivity of gadobenate dimeglumine is approximately twice that of conventional gadolinium contrast agents at all magnetic field strengths, 39 thereby permitting similar lesion conspicuity and diagnostic information with a reduced overall dose, 40,41 it has nevertheless been shown that an increased dose of 0.1 mmol/kg bodyweight leads to greater overall signal-to-noise and contrast-to-noise ratios on delayed images. 9,42 It is possible that an increased dose in this particular patient might have led to greater normal liver enhancement relative to that of the lesion, resulting in a more typical hypointense appearance. For other MR applications, a gadobenate dimeglumine dose of 0.1 mmol/kg bodyweight is either approved or widely used off-label and numerous intraindividual crossover studies with conventional gadolinium agents have reported significantly improved imaging performance with gadobenate dimeglumine at this dose The increased R1 relaxivity and thus the improved imaging performance relative to conventional agents at equivalent dose is because of weak, transient interactions of the Gd-BOPTA contrast-effective molecule of gadobenate dimeglumine with serum proteins. 50,51 Recent studies suggest that even better diagnostic performance can be achieved by modifying the relevant acquisition and/or evaluation parameters to take account of the protein interaction and increased R1 relaxivity of gadobenate dimeglumine As noted previously, 15,17,18 certain HCC may appear atypically hyperintense (Fig. 8) or isointense on hepatobiliary phase images, whereas certain FNH may appear atypically hypointense. These studies have addressed the appearance of these hepatocellular lesions in normal liver and, in the case of HCC, in cirrhotic liver. 15 Among the lesions in the present analysis, most of false-positive lesions (62 of 65) were HCC that appeared either hyperintense (n % ) or isointense (n % ) on delayed phase images. Conversely, the 13 (4.3%) false-negative lesions were all atypical FNH that appeared inhomogeneously hypointense on delayed phase images. In the case of HCC, a hyperintense or isointense appearance may be due to residual hepatocyte functionality in certain welldifferentiated lesions or due to large amounts of intralesional fatty metaplasia. 15 Conversely, certain FNH may appear atypically hypointense on delayed images because of extensive fibrosis within the lesion. 17,18 Concerning the FNH lesions included in this analysis, a Lippincott Williams & Wilkins

9 Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 4. A 32-year-old man with FNH and a history of seminoma. The lesion (arrows) appears isointense on the unenhanced T1w (A) and T2w HASTE (B) images. Strong contrast enhancement occurs during the T1w arterial phase (C) after administration of 0.05 mmol/kg gadobenate dimeglumine, with no washout during the portal-venous (D) and equilibrium (E) phases. The enhancement behavior is typical of FNH, except that a central scar is not clearly appreciated. On the delayed (2 hour) T1w hepatobiliary phase images (F), the lesion is homogeneously hyperintense against the normal liver parenchyma, indicating a lesion with functioning hepatocytes that is able to take up Gd-BOPTA. Moreover, a central scar is better appreciated. relatively high proportion (80 of 302; 26.5%) were referred for biopsy, despite the characteristic hyperintense appearance of the lesions on delayed hepatobiliary phase images. After resection, 16 lesions were further assessed histologically. Among the 80 biopsied FNH lesions were the 13 atypical hypointense lesions; the remainder were predominantly lesions with atypical features on initial dynamic phase imaging (eg, an inhomogeneous appearance, calcification, a peripheral rim, absent central scar in larger lesions). Atypical imaging findings on MR imaging are not uncommon for FNH. 55,56 However, the appearance of these lesions on delayed hepatobiliary phase images suggests that follow-up might have been an alternative approach to patient management rather than immediate biopsy. A limitation of this analysis is that the lesions included were among the more common hypervascular lesion types encountered in routine practice. Recently, a report from Marin et al 57 has shown that primary leiomyoma, which is a rare benign tumor of the liver, also appears hypointense on delayed hepatobiliary phase images after gadobenate dimeglumine administration. Given the criteria adopted for this analysis, a hypointense appearance on delayed imaging would be suggestive of a malignant lesion for which biopsy is indicated. However, although biopsy would have been unnecessary in this case, it is clearly preferable to histologically assess a suspected malignant lesion that is subsequently confirmed as benign than not to assess a lesion that is subsequently confirmed as malignant. Findings from routine practice should provide additional information on whether other benign hypervascular lesions apart from HA and leiomyoma have a hypointense appearance on delayed phase images. By analogy with HA, it may be envisaged that benign 2011 Lippincott Williams & Wilkins 233

10 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 5. A 33-year-old woman with FNH and adenomatosis. The unenhanced T1w image without fat suppression (A) reveals 1 slightly hyperintense lesion (arrow). The lesion shows signal suppression on the opposed phase T1w image (B), indicating fat within the lesion. The T2w TSE (C) image reveals 2 hyperintense lesions (arrows). On the T1w arterial phase image (D) after administration of 0.05 mmol/kg gadobenate dimeglumine, the lesions demonstrate homogeneous contrast enhancement. In this phase, a large hypervascular lesion can be seen in the fifth segment. In the T1w portal-venous phase (E), the larger lesion appears isointense, whereas the 2 smaller lesions show contrast washout (asterisk) and retention of contrast agent resulting in a hyperintense appearance (arrow), respectively. On the equilibrium phase image (F), all but 1 lesion appears isointense; only the lesion that showed contrast washout during the portal-venous phase appears hypointense. On the delayed (2 hour) T1w hepatobiliary phase images acquired without (G) and with (H) fat suppression, the 2 smaller lesions appear homogeneously hypointense, whereas the larger lesion is markedly hyperintense against the normal liver parenchyma. The 2 hypointense lesions are adenomas which lack biliary ductules, and thus are unable to take up and excrete Gd-BOPTA through the hepatobiliary route. The larger lesion is an FNH Lippincott Williams & Wilkins

11 Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 6. A 65-year-old HCV woman with a small HCC for 10 years. The unenhanced T1w image (A) reveals a small hypointense lesion in the sixth segment. The lesion appears slightly hyperintense on the T2w HASTE image (B). Strong homogeneous contrast enhancement occurs during the T1w arterial phase (C) after administration of 0.05 mmol/kg gadobenate dimeglumine, with no washout during the portal-venous (D) and equilibrium (E) phases. On the delayed (2 hour) T1w hepatobiliary phase image (F), the lesion (arrow) is homogeneously hypointense against the normal liver parenchyma, indicating a lesion that is unable to take up Gd-BOPTA Lippincott Williams & Wilkins 235

12 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 FIGURE 7. A 75-year-old woman with chronic hepatitis and HCC. The unenhanced T1w image (A) reveals a homogeneous hypointense lesion (arrow). The lesion appears hyperintense on the T2w TSE (B) and HASTE (C) images. Signs of cirrhosis are also apparent with these sequences. Strong homogeneous contrast enhancement occurs during the T1w arterial phase (D) after administration of 0.05 mmol/kg gadobenate dimeglumine, which persists into the portal-venous phase (E). During the equilibrium phase (F) and on delayed (1 hour) T1w hepatobiliary phase images acquired without (G) and with (H) fat suppression, the lesion is homogeneously hypointense against the normal liver parenchyma, indicating a lesion that is unable to take up Gd-BOPTA Lippincott Williams & Wilkins

13 Investigative Radiology Volume 46, Number 4, April 2011 TABLE 6. Sensitivity, Specificity, Accuracy, PPV and NPV for the Differentiation of Benign From Malignant Lesions on Dynamic (Portal-Venous and Equilibrium Phases: Washout) and Hepatobiliary Phase (Assuming Iso/Hyperintensity as a Sign of Benignity) Imaging Imaging Phase Lesions Sensitivity Specificity Accuracy PPV NPV Portal-venous and equilibrium phases washout % 98.7% 77.4% 98.5% 65.3% Hepatobiliary phase: all lesions % 87.6% 91.4% 85.1% 97.3% Hepatobiliary phase: lesions with homogenous % 85.4% 92.7% 89.2% 98.1% enhancement on arterial phase imaging Hepatobiliary phase: lesions with no washout on portal-venous and equilibrium phase imaging % 78.1% 90.3% 89.3% 92.9% PPV indicates positive predictive value; NPV, negative predictive value. FIGURE 8. A 77-year-old woman with chronic hepatitis and HCC. The unenhanced T1w image (A) reveals a homogeneous hypointense lesion at the dome of the liver. The lesion appears slightly hyperintense on the HASTE (B) images. Inhomogeneous contrast enhancement occurs during the T1w arterial phase (C) after administration of 0.05 mmol/kg gadobenate dimeglumine, with a washout into the portal-venous (D) and equilibrium phase (E). On delayed (2 hours) T1w hepatobiliary phase images acquired without (F) fat suppression, the lesion is inhomogeneously hyperintense against the normal liver parenchyma, indicating a lesion that is able to take up Gd-BOPTA. lesions with compromised or absent biliary ductules are more likely to appear hypointense. A second possible limitation is that alternative MR approaches such as diffusion weighted imaging (DWI) were not included in the assessment. Although DWI sequences are known to be of value for the characterization of certain hypervascular liver lesions such as hemangiomas and cysts, they are of limited use for the accurate differential diagnosis of solid hypervascular lesions These sequences were used in comparatively few subjects in our cohort, and thus their potential added value was not assessed. Future work might be directed toward evaluating the 2011 Lippincott Williams & Wilkins 237

14 Morana et al Investigative Radiology Volume 46, Number 4, April 2011 comparative benefits of DWI in combination with delayed hepatobiliary phase imaging. A final limitation is that the analysis was retrospective in nature. However, it should be borne in mind that consecutive patients were included at each center and that image assessment was based solely on lesion enhancement relative to normal liver enhancement. Nevertheless, a prospective study is perhaps warranted to confirm the conclusions of this study. Although this analysis focused solely on hypervascular liver lesions, the benefits of delayed hepatobiliary phase imaging with gadobenate dimeglumine have also been demonstrated recently for the differentiation of benign from malignant lesions that have a hypointense appearance. 28 In conclusion, this analysis demonstrates that delayed hepatobiliary phase imaging with 0.05 mmol/kg gadobenate dimeglumine is accurate for the correct differentiation of true benign hypervascular lesions from malignant and high-risk hypervascular lesions. However, it should be emphasized that delayed hepatobiliary phase imaging is not necessary for all patients referred for MRI for suspected liver lesions. Therefore, it would make little sense to perform hepatobiliary phase imaging in patients with lesions such as obvious cysts and hemangiomas, which can readily be characterized on unenhanced and postcontrast dynamic phase images alone. On the other hand, the possibility to acquire hepatobiliary phase images is potentially highly advantageous for equivocal lesions or lesions which have an atypical appearance on dynamic phase imaging. Concerning workflow and logistical issues, it should be borne in mind that approximately only 6 to 8 minutes are needed to acquire delayed hepatobiliary phase images (including patient positioning, etc.), and that delayed phase images can be acquired at any time between 1 and 3 hours after gadobenate dimeglumine injection, with later delayed acquisitions often proving beneficial compared with earlier delayed acquisitions. 17 Therefore, if additional delayed phase acquisitions are deemed necessary for a given patient, the overall workflow and timing are not greatly affected, especially because delayed acquisitions are not necessary for every patient. REFERENCES 1. Ba-Ssalamah A, Uffmann M, Saini S, et al. Clinical value of MRI liverspecific contrast agents: a tailored examination for a confident non-invasive diagnosis of focal liver lesions. Eur Radiol. 2009;19: Oliva MR, Saini S. Liver cancer imaging: role of CT, MRI, US and PET. Cancer Imaging. 2004;4(spec no A):S42 S Braga L, Guller U, Semelka RC. Modern hepatic imaging review. Surg Clin North Am. 2004;84: Elsayes KM, Leyendecker JR, Menias CO, et al. MRI characterization of 124 CT-indeterminate focal hepatic lesions: evaluation of clinical utility. HPB (Oxford). 2007;9: Hamm B, Thoeni RF, Gould RG, et al. Focal liver lesions: characterization with nonenhanced and dynamic contrast material enhanced MR imaging. Radiology. 1994;190: Morana G, Salviato E, Guarise A. Contrast agents for hepatic MRI. Cancer Imaging. 2007;7(spec no A):S24 S Morana G, Grazioli L, Testoni M, et al. Contrast agents for hepatic magnetic resonance imaging. Top Magn Reson Imaging. 2002;13: Kirchin MA, Pirovano G, Spinazzi A. Gadobenate dimeglumine (Gd- BOPTA): an overview. Invest Radiol. 1998;33: Spinazzi A, Lorusso V, Pirovano G, et al. MultiHance clinical pharmacology: biodistribution and MR enhancement of the liver. Acad Radiol. 1998;5(suppl 1): S86 S Spinazzi A, Lorusso V, Pirovano G, et al. Safety, tolerance, biodistribution and MR imaging enhancement of the liver with Gd-BOPTA: results of clinical pharmacologic and pilot imaging studies in non-patient and patient volunteers. Acad Radiol. 1999;6: Kim YK, Lee JM, Kim CS. Gadobenate dimeglumine-enhanced liver MR imaging: value of dynamic and delayed imaging for the characterization and detection of focal liver lesions. Eur Radiol. 2004;14: Kim YK, Lee JM, Kim CS, et al. Detection of liver metastases: gadobenate dimeglumine-enhanced three-dimensional dynamic phases and one-hour delayed phase MR imaging versus superparamagnetic iron oxide-enhanced MR imaging. Eur Radiol. 2005;15: Petersein J, Spinazzi A, Giovagnoni A, et al. Evaluation of the efficacy of gadobenate dimeglumine in MR imaging of focal liver lesions: a multicenter phase III clinical study. Radiology. 2000;215: Pirovano G, Vanzulli A, Marti-Bonmati L, et al. Evaluation of the accuracy of gadobenate dimeglumine-enhanced MR imaging in the detection and characterization of focal liver lesions. Am J Roentgenol. 2000;175: Grazioli L, Morana G, Caudana R, et al. Hepatocellular carcinoma: correlation between gadobenate dimeglumine-enhanced MRI and pathologic findings. Invest Radiol. 2000;35: Kim JI, Lee JM, Choi JY, et al. The value of gadobenate dimeglumineenhanced delayed phase MR imaging for characterization of hepatocellular nodules in the cirrhotic liver. Invest Radiol. 2008;43: Grazioli L, Morana G, Federle MP, et al. Focal nodular hyperplasia: morphological and functional information from MR imaging with gadobenate dimeglumine. Radiology. 2001;221: Grazioli L, Morana G, Kirchin MA, et al. Accurate differentiation of focal nodular hyperplasia from hepatic adenoma at gadobenate dimeglumine enhanced MR imaging: prospective study. Radiology. 2005;236: Marin D, Di Martino M, Guerrisi A, et al. Hepatocellular carcinoma in patients with cirrhosis: qualitative comparison of gadobenate dimeglumineenhanced MR imaging and multiphasic 64-section CT. Radiology. 2009;251: Ichikawa T, Saito K, Yoshioka N, et al. Detection and characterization of focal liver lesions: a Japanese phase III, multicenter comparison between gadoxetic acid disodium-enhanced magnetic resonance imaging and contrastenhanced computed tomography predominantly in patients with hepatocellular carcinoma and chronic liver disease. Invest Radiol. 2010;45: Sun HY, Lee JM, Shin CI, et al. Gadoxetic acid-enhanced magnetic resonance imaging for differentiating small hepatocellular carcinomas ( or 2cmin diameter) from arterial enhancing pseudolesions: special emphasis on hepatobiliary phase imaging. Invest Radiol. 2010;45: Namasivayam S, Salman K, Mittal PK, et al. Hypervascular hepatic focal lesions: spectrum of imaging features. Curr Probl Diagn Radiol. 2007;36: Chun Hsee L, McCall JL, Koea JB. Focal nodular hyperplasia: what are the indications for resection? HPB (Oxford). 2005;7: De Carlis L, Pirotta V, Rondinara GF, et al. Hepatic adenoma and focal nodular hyperplasia: diagnosis and criteria for treatment. Liver Transpl Surg. 1997;3: Herman P, Pugliese V, Machado MA, et al. Hepatic adenoma and focal nodular hyperplasia: differential diagnosis and treatment. World J Surg. 2000;24: Leese T, Farges O, Bismuth H. Liver cell adenomas: 12 years surgical experience from a specialist hepato-biliary unit. Ann Surg. 1988;208: Semelka RC, Brown ED, Ascher SM, et al. Hepatic hemangiomas: a multiinstitutional study of appearance on T2-weighted and serial gadoliniumenhanced gradient-echo MR images. Radiology. 1994;192: Hwang HS, Kim SH, Jeon TY, et al. Hypointense hepatic lesions depicted on gadobenate dimeglumine-enhanced three-hour delayed hepatobiliary-phase MR imaging: differentiation between benignancy and malignancy. Korean J Radiol. 2009;10: Schneider G, Grazioli L, Saini SImaging of benign focal liver lesions. In: MRI of the Liver: Imaging Techniques, Contrast Enhancement, Differential Diagnosis. 2nd ed. Milan, Italy: Springer-Verlag Italia; 2006: Grazioli L, Federle MP, Ichikawa T, et al. Liver adenomatosis: clinical, histopathologic, and imaging findings in 15 patients. Radiology. 2000;216: Chiche L, Dao T, Salame E, et al. Liver adenomatosis: reappraisal, diagnosis, and surgical management: eight new cases and review of the literature. Ann Surg. 2000;231: Namasivayam S, Martin DR, Saini S. Imaging of liver metastases: MRI. Cancer Imaging. 2007;7: Murakami T, Nakamura H, Tsuda K, et al. Contrast-enhanced MR imaging of intrahepatic cholangiocarcinoma: pathologic correlation study. J Magn Reson Imaging. 1995;5: Mahfouz AE, Hamm B, Wolf KJ. Peripheral washout: a sign of malignancy on dynamic gadolinium-enhanced MR images of focal liver lesions. Radiology. 1994;190: Lippincott Williams & Wilkins