The role of dynamic subtraction MRI in detection of hepatocellular carcinoma

Transcription

1 Diagn Interv Radiol 2008; 14: Turkish Soiety of Radiology 2008 ABDOMINAL IMAGING ORIGINAL ARTICLE The role of dynami subtration MRI in detetion of hepatoellular arinoma Mustafa Seçil, Funda Obuz, Canan Altay, Ömür Genel, Enis İği, Özgül Sağol, Oğuz Dile PURPOSE The aim of this study was to investigate the role of dynami subtration magneti resonane imaging (MRI) in detetion of hepatoellular arinoma (HCC) in patients with irrhosis. MATERIALS AND METHODS MR images of 32 patients with irrhosis were retrospetively investigated. The standard sequene images of the patients were evaluated for detetion of HCC, and then the subtrated images of dynami ontrast-enhaned series were evaluated. The sensitivity, speifiity, and auray of standard tehnique and additional use of subtrated images were investigated. RESULTS In detetion of at least one HCC, standard protool was useful in 14 of 17 (82.3%) patients; by additional use of subtration imaging all 17 (100%) patients with HCC were deteted. For detetion of the orret number of HCC tumors, standard protool was found to have 61.5% sensitivity, 78.9% speifiity, and 71.8% auray. The use of subtrated images inreased the sensitivity to 85.7%, speifiity to 83.3%, and auray to 84.3%. CONCLUSION Subtration is a simple automati proedure that is ommonly available in most MRI systems. The use of subtration of dynami ontrast-enhaned series failitates the detetion of HCC in disorganized arhiteture of irrhoti livers. Key words: irrhosis arinoma, hepatoellular magneti resonane imaging From the Departments of Radiology (M.S. mustafa.seil@deu. edu.tr, F.O., C.A., Ö.G., E.İ., O.D.), and Pathology (Ö.S.), Dokuz Eylül University Shool of Mediine, İzmir, Turkey. Reeived 11 April 2007; revision requested 2 August 2007; revision reeived 9 August 2007; aepted 19 September C irrhosis is a diffuse liver disease haraterized by progressive parenhymal damage and nodular regeneration (1 3). Hepatoellular arinoma (HCC) is a neoplasm that usually arises in a irrhoti liver by a multistep arinogenesis proess (4). Reent studies have shown that in patients with irrhosis and early stage HCC, liver transplantation offers the best hane for long-term survival (2, 5, 6). Therefore, early detetion of HCC and aurate assessment of tumor burden are ruial to suessful treatment planning and long-term survival. Magneti resonane imaging (MRI) plays a prominent role in the evaluation of irrhosis and sreening for early HCC (5). Gadolinium-enhaned MRI with multiple phases of aquisition improves the detetion of HCCs (1, 7). However, the determination of ontrast enhanement is not always easy to aomplish for hyperintense lesions on arterial phase dynami images (8). Subtration of unenhaned images from gadolinium-enhaned images has been pursued in an attempt to maximize the qualitative reognition of lesion enhanement (8, 9). The aim of this study was to investigate the role of subtration MRI in detetion of HCC in patients with irrhosis. Materials and methods Patients The study group of this retrospetive investigation was omposed of patients with irrhosis who had liver MRI in piture arhiving and ommuniation system (PACS) arhive. The MRI of the patients had been performed to evaluate the severity of irrhosis or portal hypertension, sreening for hepati lesions suspeted with other imaging modalities. Two investigators (C.A., O.G.) reviewed the linial reords and the images of the patients to determine the inlusion of the patients in the study. Inlusion riteria of the study were: histopathologial diagnosis of irrhosis, linial and MRI follow-up >1 year, and presene of a omplete series of standard liver MR images aording to the protool of our institute, inluding multiphasi dynami ontrast enhaned MRI series with high-quality breath holding. MRI tehnique MRI was performed with a 1.5-T unit (Intera, software version 8.1; Philips Medial Systems, Eindhoven, The Netherlands) by using a phased-array oil. The standard liver MR protool of our institute inluded T2-weighted fat-saturated spin eho (SE) (TR/TE/FA, 1600/70/ 90 ), heavy T2-weighted single shot (1312/325/90 ), T1-weighted inphase gradient eho (GRE) (196/4.6/80 ), T1-weighted opposed-phase GRE (253/6.9/80 ), T1-weighted water seletive (WATS) fat-saturated multiphasi ontrast-enhaned dynami MRI (272/6.9/70 ) in transverse plane, and delayed post-ontrast T1-weighted WATS in oronal 200

2 plane. Multiphasi ontrast enhaned dynami series were obtained just before and during the rapid bolus intravenous injetion of 0.1 mmol gadopentetate dimeglumine per kilogram of body weight while the patient was in the bore of the magnet. The imaging timing of the dynami series inluded pre-ontrast, arterial, portal, and equilibrium phases of the liver. Subtration of multiphasi ontrast enhaned dynami series was automatially aquired by the software of MR mahine. The software provided a new series by image-by-image subtration of pre-ontrast series from eah post-ontrast series (arterial, portal, and equilibrium) of eah patient. Image analysis Image analysis was performed independently by two other investigators (M.S., F.O.) who were experiened in liver MRI. The investigators were unaware of the linial ondition of the patients; all patient data were hidden during the image analysis. The investigators evaluated the standard protool MR images first and noted their findings on the presene and number of HCC lesions. Later in the same session, subtrated images of patients were examined for the same purpose, and deisions were noted. In ases of onflit in the deisions of the investigators, the images were reevaluated jointly, and a final deision was reahed by onsensus. was performed, and basi statistial parameters of sensitivity, speifiity, positive and negative preditive values, and the auray of eah method were alulated. Results Thirty-two patients met the inlusion riteria of the study; 15 did not have HCC, and 17 had 1 HCC lesion. In detetion of 1 HCC lesion, standard protool was useful in 14 of 17 (82.3%) of the patients; by inluding subtration imaging, all of 17 (100%) patients with HCC were deteted. Among the patients with HCC, final diagnosis showed that 12 patients had a single tumor, 3 had two tumors, 1 had three tumors, and 1 had five tumors. A omparison of methods for detetion of HCC tumors is presented in Table 1. The patients who were underdiagnosed (false negative) by subtration imaging were the patients who were also underdiagnosed by standard protool. In one patient with 5 HCC lesions and another patient with 3 HCC lesions in whom the diagnoses were reahed by histopathologial evaluation of the explanted liver, eah method underdiagnosed the number of tumors beause the missed lesions were <1 m in diameter. The other underdiagnoses by standard protool in 3 patients were found to originate from equivalent intensity of lesions on T2 and high intensity on T1-weighted images when ompared to the liver parenhyma (Fig. 1). Overdiagnoses of tumors (false positive) by both methods ourred in 2 patients. One patient with HCC was found to have a hamartoma, whih was misdiagnosed as a seond HCC tumor. In the other patient, a faintly enhaning nodular area was assessed as an HCC by both methods, but histopathologial diagnosis did not support this diagnosis. Standard protool alone made an overdiagnosis in one patient, aused by a dysplasti nodule whih aused diffiulty in estimation of ontrast enhanement on dynami series of standard protool beause of hyperintensity on T1-weighted images (Fig. 2). Standard + subtration imaging yielded overdiagnosis in one patient, in whom a dysplasti nodule was diagnosed as HCC subsequent to subtration misregistration of the images (Fig. 3). The true negative rate did not show any differene between the methods; however, true positive rates were found to inrease and false negative and false positive rates were found to derease by the use of subtration imaging. The basi statistial results for detetion of the number of HCC tumors in patient-based analysis are presented in Table 2. Final diagnosis The final HCC diagnoses of the patients were reahed by histopathologial examination of explanted liver (n = 4) or reseted speimen (n = 1) in operable ases. In inoperable ases, perutaneous biopsy (n = 2) or hemoembolization and lipiodol CT (n = 5), plus 1 year MR follow-up of patients were used for final diagnoses. The absene of HCC was onfirmed by histopathologial examination of explanted liver (n = 1) and 1 year linial follow-up. In patients with HCC, the absene of additional HCC lesions in other liver areas was onfirmed by linial evaluation and 1-year MRI follow-up. Statistial analysis A patient-based analysis of the results for standard protool and for subsequent use of subtration images Volume 14 Issue 4 Table 1. Patient-based analysis by both methods for detetion of hepatoellular arinoma Patients (n) Standard protool Standard protool + subtration True positive 9 12 False positive (overdiagnosis) 5 3 False negative (underdiagnosis) 4 2 True negative Table 2. Statistial results of methods for detetion of hepatoellular arinoma Standard protool Standard protool + subtration Sensitivity 61.5% 85.7% Speifiity 78.9% 83.3% Positive preditive value 66.6% 80.0% Negative preditive value 75.0% 88.2% Auray 71.8% 84.3% Subtration MRI in hepatoellular arinoma 201

3 a b d Figure 1. a d. A 51-year-old man with irrhosis. Axial T2-weighted MR image (a) shows heterogeneous parenhyma with multiple nodules and a hyperintense lesion in the posterior setor (arrow). Axial pre-ontrast T1-weighted MR image (b) shows multiple hyperintense nodular lesions (arrows). On post-ontrast axial T1-weighted MR image (), the hyperintense lesions beame isointense with the parenhyma exept for the hepatoellular arinoma (HCC) lesion in the posterior setor (arrow). Subtrated MR image (d) demonstrates a lesion in a different loation (arrow) from the other nodular lesions, whih was proven to be an HCC. a b Figure 2. a. A 59-year-old man with irrhosis. Axial pre-ontrast T1-weighted MR image (a), shows two hyperintense nodules (long and short arrows). On axial post-ontrast MR image (b), one of these lesions is isointense with the hepati parenhyma (long arrow), and the other is hypointense (short arrow); in both, the estimation of ontrast enhanement amount is diffiult. By subtration (), both lesions appear nonenhaning lesions ompatible with dysplasti nodules (long and short arrows). Both lesions were proven to be dysplasti nodules. Disussion The main purpose of imaging in irrhosis is to identify HCC. The distorted arhiteture of liver parenhyma, nodular regeneration, and signal intensity variability of the nodules ause diffiulties in detetion of HCC (8). Contrast-enhaned multiphasi dynami sequenes have beome a standard of liver MRI in irrhosis. Arterial phase enhanement after gadolinium administration has been proposed as the 202 Deember 2008 Diagnosti and Interventional Radiology Seçil et al.

4 a ter of the lesions on T1-weighted images. The amount of enhanement in hyperintense nodular lesions reates diffiulty in evaluation of T1-weighted images (8). However, in subtration imaging, the baseline hyperintensity of a lesion is erased by subtration, and only the hyperintensity due to ontrast enhanement remains. Hene, subtration images failitate the ability to see the ontrast enhanement of a lesion. The use of standard + subtration imaging was also better than standard protool alone for detetion of the orret number of HCC lesions. All patients who were underdiagnosed by subtration imaging were also underdiagnosed by standard protool alone. The main reason for missing the existing tumors (false negatives) in these patients was the small size of the missed tumors. The limitation of MRI in detetion of small size HCCs has been demonstrated (1). Moreover, these tumors may have low vasularity and may not show ontrast enhanement (1, 2). Likely for these reasons, standard protool and standard + subtration imaging provided the same number of underdiagnosed tumors in these patients. Underdiagnosis of three other patients using standard protool alone originated from equivalent intensity of lesions on T2- and high intensity on T1-weighted images when ompared to the liver parenhyma. Previous studies have reported that the suess of subtration tehnique deb Fig. 3. a. False positive result of subtration imaging in a 64-year-old woman. Axial pre-ontrast T1-weighted (a), post-ontrast T1-weighted (b), and subtration () MR images. Misregistration of the hyperintense area has aused a false enhaning nodular lesion that was aepted as hepatoellular arinoma by subtration (arrows in a ). The nodular lesion was proven to be a dysplasti nodule. most sensitive sign for the detetion of HCCs (10, 11). It is diffiult, however, to visually detet enhanement generated by gadolinium-helate administration for nodules with higher signal intensity than hepati parenhyma. Although it has been used for years in MRI of breast and MR angiography, subtration MRI in detetion of HCC in irrhosis is a fairly new onept, and only a few investigations exist in the literature (8, 9, 12, 13). One of these artiles was a pitorial essay of the potential use of subtration MRI for the liver (9); this was followed by an original researh artile on HCC detetion performed by the same authors (8). Previous researh on HCC detetion used a two-step investigation that inluded first the tehnial feasibility of subtration and then the haraterization of hyperintense lesions by onventional versus subtration images of post-ontrast T1-weighted series. Our study has been designed as a retrospetive investigation of the images of irrhoti patients with arhived images taken more than a year after the original images. Tehnial feasibility of the subtration method was not taken into onsideration in our study by elimination of patients with low-quality breath-hold images that may ause subtration artifat. The main onern of our study was to investigate the potential benefit of subtration imaging in addition to the onventional sequenes. Aording to the results of our study, inluding use of subtration imaging yielded inreased sensitivity, speifiity, and auray rates positive and negative preditive values ompared to the use of standard protool alone. In assessing absene of HCC, both methods obtained the same results and orretly determined tumor absene in all 15 patients. However, for detetion of at least one HCC lesion, standard + subtration imaging was superior to the standard protool. In 3 patients, HCCs were overlooked by standard protool images and deteted by standard + subtration imaging. The main reason for this was the hyperintense hara- Volume 14 Issue 4 Subtration MRI in hepatoellular arinoma 203

5 pends on the degree of misregistration artifat between the non-enhaned and enhaned soure images, as well as the size and loation of the nodules (8). In our study, we tried to eliminate the risk of misdiagnosis aused by misregistration artifat by exluding patients with suboptimal breath-holding. However, a misregistered peripheral nodule at the dome of the liver aused a false positive result in one of our patients. Peripheral lesions at the liver dome have been reported to be at partiular risk for misregistration artifat, despite efforts to optimize through use of oregistrations (8). The reported sensitivity rates of MRI for detetion of HCC vary widely, from 55% to 100% (2, 5 8, 12, 14 19). There have been various attempts to inrease the detetion rate of HCC with the use of new ontrast agents or ombined use of ontrast materials (2, 15, 17, 20). Subtration imaging is a no-ost method that is simply aquired by most of the MRI devies by automati subtration of pre-ontrast images from post-ontrast dynami T1-weighted series. Our results suggest that further improvement of detetion rate of HCC lesions by MRI may be ahieved by inlusion of subtration imaging in standard protool. This study has several limitations. First, there was no onsistent gold standard for final diagnoses in our study. The ideal standard of histopathologial evaluation of explanted liver ould not be ahieved in all patients. We tried to overome this limitation by inluding patients who had available follow-up images taken more than one year after the initial images. This, however, led to the seond limitation of the study, whih was the relatively low number of patients. Third, the study was a retrospetive investigation of arhived images, whih limited the use of newer pulse sequene designs suh as thinner setion three-dimensional imaging. Finally, the evaluation of both methods in the same session might seem to have introdued bias, but our aim was to determine the value of subsequent evaluation of subtration imaging, whih is ommonly used in daily pratie. In onlusion, subtration is a simple automati proedure that is ommonly available in most MRI mahines and the use of subtration of dynami ontrast enhaned series is a helpful, no-ost tool that improves detetion of HCC. Referenes 1. Ward J, Robinson PJ. How to detet hepatoellular arinoma in irrhosis. Eur Radiol 2002; 12: Ward J, Guthrie JA, Sott DJ, et al. Hepatoellular arinoma in the irrhoti liver: double-ontrast MR imaging for diagnosis. Radiology 2000; 216: Branatelli G, Federle MP, Ambrosini R, et al. Cirrhosis: CT and MR imaging evaluation. Eur J Radiol 2007; 61: Efremidis SC, Hytiroglou P. The multistep proess of hepatoarinogenesis in irrhosis with imaging orrelation. Eur Radiol 2002; 12: Heht EM, Holland AE, Israel GM, et al. Hepatoellular arinoma in the irrhoti liver: gadolinium-enhaned 3D T1-weighted MR imaging as a stand-alone sequene for diagnosis. Radiology 2006; 239: Krinsky GA, Lee VS, Theise ND, et al. Hepatoellular arinoma and dysplasti nodules in patients with irrhosis: prospetive diagnosis with MR imaging and explantation orrelation. Radiology 2001; 219: Shimizu A, Ito K, Koike S, Fujita T, Shimizu K, Matsunaga N. Cirrhosis or hroni hepatitis: evaluation of small (<or=2-m) early-enhaning hepati lesions with serial ontrast-enhaned dynami MR imaging. Radiology 2003; 226: Yu JS, Kim YH, Rofsky NM. Dynami subtration magneti resonane imaging of irrhoti liver: assessment of high signal intensity lesions on nonenhaned T1- weighted images. J Comput Assist Tomogr 2005; 29: Yu JS, Rofsky NM. Dynami subtration MR imaging of the liver: advantages and pitfalls. AJR Am J Roentgenol 2003; 180: Bhartia B, Ward J, Guthrie JA, Robinson PJ. Hepatoellular arinoma in irrhoti livers: double-ontrast thin-setion MR imaging with pathologi orrelation of explanted tissue. AJR Am J Roentgenol 2003; 180: Ito K. Hepatoellular arinoma: onventional MRI findings inluding gadoliniumenhaned dynami imaging. Eur J Radiol 2006;58: Soyer P, Spelle L, Gouhiri MH, et al. Gadolinium helate-enhaned subtration spoiled gradient-realled eho MR imaging of hepati tumors. AJR Am J Roentgenol 1999; 172: Savi G, Yazii Z, Sahin N, Akgoz S, Tunel E. Value of hemial shift subtration MRI in haraterization of adrenal masses. AJR Am J Roentgenol 2006; 186: Karadeniz-Bilgili MY, Braga L, Birhard KR, et al. Hepatoellular arinoma missed on gadolinium enhaned MR imaging, disovered in liver explants: retrospetive evaluation. J Magn Reson Imaging 2006; 23: Hashimoto M, Eto M, Akishita M, et al. Correlation between flow-mediated vasodilatation of the brahial artery and intima-media thikness in the arotid artery in men. Arteriosler Thromb Vas Biol 1999; 19: Kwak HS, Lee JM, Kim CS. Preoperative detetion of hepatoellular arinoma: omparison of ombined ontrast-enhaned MR imaging and ombined CT during arterial portography and CT hepati arteriography. Eur Radiol 2004; 14: Kwak HS, Lee JM, Kim YK, Lee YH, Kim CS. Detetion of hepatoellular arinoma: omparison of ferumoxides-enhaned and gadolinium-enhaned dynami three-dimensional volume interpolated breath-hold MR imaging. Eur Radiol 2005; 15: Lutz AM, Willmann JK, Goepfert K, Marinek B, Weishaupt D. Hepatoellular arinoma in irrhosis: enhanement patterns at dynami gadolinium- and superparamagneti iron oxide-enhaned T1-weighted MR imaging. Radiology 2005; 237: Noguhi Y, Murakami T, Kim T, et al. Detetion of hepatoellular arinoma: omparison of dynami MR imaging with dynami double arterial phase helial CT. AJR Am J Roentgenol 2003; 180: Simon G, Link TM, Wortler K, et al. Detetion of hepatoellular arinoma: omparison of Gd-DTPA- and ferumoxides-enhaned MR imaging. Eur Radiol 2005; 15: Deember 2008 Diagnosti and Interventional Radiology Seçil et al.