| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology and Center for Imaging Science, Sungkyunkwan University School of Medicine, 50 Ilwon-Dong, Kangnam-Ku, Seoul 135-710, Korea. Received November 14, 2001; revision requested January 24, 2002; revision received March 5; accepted April 16. Address correspondence to J.H.L. (e-mail: jhlim@smc.samsung.co.kr).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Seventy consecutive patients with a total of 79 HCC nodules underwent ferumoxides-enhanced MR imaging and triple-phase helical CT before surgery. The diagnosis of HCC was established by means of pathologic examination after surgical resection in all patients. MR images obtained with all sequences and triple-phase helical CT images were reviewed independently by three radiologists on a segment-by-segment basis. Accuracy for diagnosis of HCC was assessed by applying receiver operating characteristic (ROC) analysis to observations of 78 hepatic segments with at least one HCC nodule and 70 segments without HCC.
RESULTS: The diagnostic accuracy of findings at ferumoxides-enhanced MR imaging (with mean area-under-the-ROC-curve [Az] values for the three observers of 0.986, 0.979, and 0.980) was significantly higher (P < .001) than that of findings at triple-phase helical CT (with mean Az values for the three observers of 0.945, 0.948, and 0.964). The mean sensitivity of MR imaging (95%, 222 of 234 segments) was also significantly higher than that of triple-phase helical CT (88%, 205 of 234 segments) (P = .001, McNemar test). The mean specificity was 97% (261 of 270 segments) for MR imaging and 98% (264 of 270 segments) for CT, but this difference was not significant (P = .754, McNemar test).
CONCLUSION: Ferumoxides-enhanced MR imaging is superior to triple-phase helical CT for the preoperative depiction of HCC.
© RSNA, 2003
Index terms: Computed tomography (CT), phase imaging, 761.12114 • Iron • Liver neoplasms, CT, 761.12114 • Liver neoplasms, MR, 761.12143 • Magnetic resonance (MR), contrast media, 761.12143 • Receiver operating characteristic (ROC) curve
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The usefulness of ferumoxides-enhanced magnetic resonance (MR) imaging has been reported for detection of hepatic tumors, including HCC (6–8). Ferumoxides particles coated with dextran and administered intravenously are ingested by macrophages of the reticuloendothelial system, including the Kupffer cells of the liver; this results in signal intensity loss in normal liver tissue (9). Kupffer cells are lacking or markedly decreased in number in most malignant tumors of the liver and therefore retain their signal intensity after ferumoxides administration, resulting in the improvement of lesion-to-liver contrast (9–13). Some investigators have suggested that ferumoxides-enhanced MR imaging is more accurate than dual-phase helical CT for the detection of focal liver lesions (14–16). The purpose of this study was to compare ferumoxides-enhanced MR imaging with triple-phase helical CT for the preoperative depiction of HCC.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
-fetoprotein levels. Among these patients, a total of 70 consecutive individuals (59 men and 11 women; age range, 30–76 years; mean age, 52 years) who had pathologically proved HCC at hepatic resection surgery were selected for this study. The criterion for selecting a patient for hepatic resection surgery was the presence of a tumor or a number of tumors of any size limited to one lobe of a liver with Child-Pugh class A function. The following surgeries were performed within 3 weeks after the imaging studies: segmentectomy (n = 35), lobectomy (n = 33), extended right lobectomy (n = 1), and liver transplantation (n = 1). A total of 175 hepatic segments were resected. Surgeons confirmed the absence of tumor in any nodule thought to be suspicious at imaging by means of palpation and intraoperative US. The resected specimens were serially sliced with 5–10-mm thicknesses in the transverse or coronal plane, depending on the location of a nodule.
Seventy-nine HCC nodules in 78 segments were confirmed at pathologic examination. HCC nodules ranged from 0.6 to 12.5 cm in diameter (mean, 3.9 cm). In 45 (64%) of the 70 patients, liver cirrhosis was confirmed at histopathologic examination of the resected portion of the hepatic parenchyma. Eight of the 70 patients had hepatitis B virus–related chronic active hepatitis, and two of the 70 patients had hepatitis C virus–related chronic persistent hepatitis. The liver was normal histopathologically, with no hepatitis or cirrhosis, in 15 of the 70 patients. This study was approved by the institutional review board of our hospital, and written informed consent was obtained from all patients.
Triple-Phase Helical CT
CT was performed with a helical scanner (HiSpeed Advantage; GE Medical Systems, Milwaukee, Wis). Scanning parameters were 120 kVp, 180 mAs, 7-mm section collimation, and 7-mm/sec table speed during a single-breath-hold helical acquisition of 25–30 seconds (depending on liver size). Images were obtained in a craniocaudal direction and were reconstructed every 7 mm to provide contiguous sections. The hepatic arterial phase, portal venous phase, and delayed phase images were obtained with delays of 30, 60, and 180 seconds, respectively, after injection of 100 mL of nonionic iodinated contrast material (iopamidol, Iopamiro 300; Bracco, Milan, Italy) through the antecubital vein at a rate of 3 mL/sec.
Ferumoxides-enhanced MR Imaging
MR imaging was performed in all patients within 10 days after triple-phase helical CT with a 1.5-T MR unit (Horizon; GE Medical Systems) and a phased-array multicoil as a receiver coil. Ferumoxides solution (Feridex IV; Advanced Magnetics, Cambridge, Mass), at a dose of 15 µmol per kilogram of body weight, was diluted with 100 mL of a 5% glucose solution and administered intravenously through an in-line 5-µm specific filter for approximately 30 minutes. MR imaging began about 30 minutes after the end of ferumoxides administration.
The MR protocol included a fat-suppressed respiratory-triggered fast spin-echo sequence with two echo times (repetition time msec/echo times msec, 3,333–8,571/18 and 90–117; an echo train length of 10–18; two signals acquired; a 256 x 256 matrix; and bandwidth of 120 Hz per pixel), a gradient-recalled echo (GRE) acquisition in the steady state (216/20, 20° flip angle, one signal acquired, and bandwidth of 83.3 Hz per pixel), a T2*-weighted fast multiplanar GRE acquisition in the steady state (130/8.4–9.5, 30° flip angle, and bandwidth of 60 Hz per pixel), a T2*-weighted fast multiplanar spoiled GRE sequence (130/8.4–9.5, 30° flip angle), and a breath-hold in-phase T1-weighted fast multiplanar spoiled GRE sequence (200/4.2, 90° flip angle, and 256 x 160 matrix).
Twenty transverse images were obtained during each pulse sequence with a section thickness of 6–8 mm and a 2-mm intersection gap. Saturation bands superior and inferior to the imaging volume were applied in all sequences to minimize motion artifact. Fat saturation was used in intermediate-weighted and T2-weighted fast spin-echo sequences. During the T2*-weighted fast multiplanar GRE acquisition in the steady state and the fast multiplanar spoiled GRE sequence, multiple sections were obtained in an interleaved fashion, and the whole liver was imaged with 20 sections during two breath holds.
Resected Liver Segments
A total of 175 segments were resected in the 70 patients. However, in seven instances, two segments were completely occupied by a single HCC nodule and were considered to be a single segment. Therefore, a total of 168 segments (78 with at least one HCC nodule and 90 without HCC) were ultimately included in subsequent receiver operating characteristic (ROC) analysis. Among these 168 segments, 74 contained only one HCC nodule; one, two HCC nodules; one, one HCC and three dysplastic nodules; two, one HCC and one dysplastic nodule; and three, only one dysplastic nodule. There were no other focal lesions such as cysts or hemangiomas in any resected segments.
Image Analysis
Images from all ferumoxides-enhanced MR imaging sequences and triple-phase helical CT images were evaluated independently and in blinded fashion by three gastrointestinal radiologists (J.H.L., S.H.K., D.C.). The MR and CT images were reviewed in individual sessions separated by 3-week intervals. All images were evaluated at a 2,000 x 2,000 picture archiving and communication system (PACS; GE Medical Systems Integrated Imaging Solutions, Mount Prospect, Ill) monitor, with adjustment of the optimal window setting in each case. Narrow window settings had been used in all patients because the reviewer could adjust all parameters of window width and level at the PACS monitor. The image review was conducted on a segment-by-segment basis, but the entire liver was reviewed in every patient. The study coordinator (B.K.K.) assigned numbers to all hepatic segments according to the Couinaud numbering system.
Each observer recorded both the location (ie, the Couinaud segment number) and the size of each focal lesion and stated whether he or she was able to ascertain the presence or absence of HCC. On CT images, nodules showing enhancement during the hepatic arterial phase and a lack of portal venous supply (3–5,17) were regarded as HCC nodules. Specifically, nodules that showed homogeneous or variegated enhancement and visible intratumoral arteries on hepatic arterial phase images, isoattenuation or low attenuation on portal venous phase images, and low attenuation on delayed phase images, and nodules of mixed attenuation on hepatic arterial and portal venous phase images and low attenuation on delayed phase images were considered to be HCC nodules. In addition, nodules seen to have discrete capsules on hepatic arterial and delayed phase images (18), nodules with a mosaic appearance on portal venous and delayed phase images (18), and nodules larger than 2 cm that showed predominantly low attenuation during all three phases (4) were regarded as HCC.
On ferumoxides-enhanced MR images, any visible nodule of high signal intensity was considered to represent HCC. Each observer assigned one of the following five confidence levels to his or her observations: 1, HCC definitely absent; 2, HCC probably absent; 3, HCC possibly present; 4, HCC probably present; and 5, HCC definitely present. When a lesion invaded two or more segments, the observers were asked to evaluate only the segment that was mainly involved and to determine the probability of another lesion existing in the other segments.
Statistical Analysis
For each imaging modality, a binomial ROC curve was fitted to each observer’s confidence rating data by means of a maximum-likelihood estimation. The diagnostic accuracy of each imaging modality was determined by calculating the area under each observer-specific binomial ROC curve (ie, the area under the ROC curve [Az] index). Composite ROC curves for the combined performance of all observers were obtained by applying the maximum-likelihood curve-fitting algorithm to the pooled data of the three observers for each imaging modality. The number of HCC nodules correctly classified as possibly present (a score of 3), probably present (a score of 4), or definitely present (a score of 5) by each observer was regarded as the number of correctly diagnosed HCC nodules; these criteria are currently used in clinical practice in our hospital.
Sensitivity and specificity were calculated for each observer and for each imaging modality, and the statistical significance of any difference between MR and CT results was assessed by using the McNemar test. The positive and negative predictive values and false-positive rates for the depiction of HCCs were also calculated for each imaging modality. A P value < .05 was considered to indicate a statistically significant difference. Interobserver agreement for the detection of HCC with each imaging modality was assessed with 
statistics. > .80 were considered to indicate excellent agreement.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Four of the six false-positive CT results were attributed to the fact that two of eight dysplastic nodules were slightly hypoattenuating at triple-phase helical CT (one was hypoattenuating on images from all three phases, and the other was isoattenuating on arterial phase images and hypoattenuating on portal and delayed phase images), while the remaining six dysplastic nodules were isoattenuating. The other two false-positive CT results were due to the likely presence of arterioportal shunts.
Overall, 222 true-positive findings and nine false-positive findings were recorded at MR, with a resulting false-positive rate of 3% (nine of 270 segments), and 205 true-positive findings and six false-positive findings were recorded at CT, with a resulting false-positive rate of 2% (six of 270 segments). MR findings had a positive predictive value of 96% (222 of 231 segments) and CT findings 97% (205 of 211 segments); MR findings had a negative predictive value of 97% (261 of 273 segments) and CT findings 92% (270 of 293 segments).
The values among the three observers showed excellent agreement at both imaging modalities (Table 3).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Ferumoxides-enhanced MR imaging has been evaluated in comparison with other imaging modalities, such as abdominal US, contrast material–enhanced CT, CT during arterial portography, dynamic gadolinium-enhanced MR imaging, and unenhanced MR imaging (6,8,10,11,14–16,21,22). In all but one (10) of these studies, ferumoxides-enhanced MR imaging appeared to be superior to other imaging modalities for the depiction of hepatic tumors. However, most of these studies had limitations, including a lack of histopathologic confirmation and heterogeneity in the histologic types of hepatic tumors. In our study, the only patients included were those who were confirmed to have HCC at histopathologic evaluation after hepatic resection.
With pathologic findings in the resected liver specimens as the standard of reference, the results of our study revealed that ferumoxides-enhanced MR imaging was superior to triple-phase helical CT for depiction of HCC. The mean sensitivity of ferumoxides-enhanced MR imaging (95%, 222 of 234 segments) was significantly higher than that of triple-phase helical CT (88%, 205 of 234 segments; P = .001). The specificities of MR and CT were 97% and 98%, respectively, and the difference between them was not significant. Bluemke et al (16) reported that the specificity of dual-phase helical CT was superior to that of ferumoxides-enhanced MR imaging in the detection of hepatic lesions. However, in the study of Bluemke et al (10), all selected patients had hepatic metastases, not HCC nodules.
Our study did not use explanted liver as the ultimate diagnostic standard (we instead relied on findings in the resected specimens for pathologic proof), and this factor probably led to higher sensitivities. Peterson et al (23) used pathologic findings in 430 explanted livers as the diagnostic standard and reported that triphasic CT (CT with precontrast, hepatic arterial, and portal venous phases) depicted HCC in 26 (59%) of 44 patients prospectively and 30 (68%) of 44 patients retrospectively. On the basis of findings in explanted livers, Lim et al (24) found that triphasic helical CT (CT with hepatic arterial, portal venous, and equilibrium phases) had a per-patient sensitivity for depiction of HCC of 80% (ie, CT depicted HCC in 12 of 15 affected patients) and a detection sensitivity of 71% (ie, CT depicted 15 of 21 HCC nodules). After another study in which findings in explanted liver were used as the diagnostic standard, Krinsky et al (25) reported that dynamic gadolinium-enhanced MR images depicted 11 of 20 hepatic tumors in 71 patients after transplantation, for a sensitivity of 55% (11 of 20 tumors) and a specificity of 86% (60 of 70 patients).
Although pathologic findings in explanted liver constitute the best standard of reference, the group of patients evaluated in the study of Krinsky et al (25) who had undergone transplantation represented patients with late-stage or end-stage cirrhosis; therefore, they do not represent the typical group of patients encountered in day-to-day clinical practice. In advanced and late stages of liver cirrhosis, the detectability of HCC is relatively low because of extensive fibrosis and distortion of liver parenchyma (23–25).
In contrast to the results of the studies described above, Tang et al (10), after a prospective study of 53 patients with cirrhosis, reported that the overall sensitivity of dynamic gadolinium-enhanced MR imaging was 94% (97 of 103 lesions) and that of ferumoxides-enhanced T2*-weighted MR imaging was 78% (80 of 103 lesions). After another more recent study that incorporated a segment-by-segment analysis, Matsuo et al (26) reported results of ferumoxides-enhanced MR imaging that were not as good as our results: The sensitivity in that study was 62% (150 of 234 segments), and the Az value was 0.807. However, the studies of Tang et al (10) and Matsuo et al (26) included cases without histologically proved malignancy, and the majority of their patients had severe cirrhosis (38 of 53 patients) (10) or a degree of cirrhosis that was not described (26). It is well known that the effectiveness of ferumoxides-enhanced MR imaging is limited in patients with severe cirrhosis (10,27).
We assessed the accuracy of CT and MR imaging in the depiction of HCC in patients whose liver function was relatively good, enabling them to sustain surgery, rather than in patients with end-stage cirrhosis. Therefore, the results of our study are probably better than the results that will be achieved in clinical practice, in which patients with advanced cirrhosis are often encountered. This inherent weakness could not be avoided with our study policy of confirmation of presence or absence of HCC by means of pathologic proof. Without sampling the entire liver, the true sensitivity and specificity of findings at any given imaging modality remain unknown.
If we were to have included liver segments without tumors at the time of observation, assessed the presence of HCC during a long period of interval follow-up of the remaining liver segments, and used findings at follow-up as a reference, we could have excluded patient selection bias. However, this procedure would not have represented a valid standard of reference, because not all small HCCs will be depicted at imaging examinations. Furthermore, new HCC nodules can develop in patients with liver cirrhosis after a certain time. Therefore, with current methods, it is impossible to determine the true sensitivity and specificity of an imaging modality for the general patient population seen in day-to-day clinical practice. For the evaluation of the accuracy of a test for the detection of HCC, patients having relatively good liver function—for example, candidates for hepatic resection surgery—should be grouped differently from patients with late-stage cirrhosis—for example, candidates for transplantation. In fact, the detectability of HCC in patients who have undergone hepatic resection surgery is not as low as it is in patients who have undergone liver transplantation.
There is special concern regarding the use of ferumoxides in patients with severe cirrhosis. Scarring, regeneration, inflammation, and shunting in cirrhotic livers can reduce the hepatic uptake of ferumoxides and result in signal intensity heterogeneity in the liver (Fig 3), thereby limiting the effectiveness of postcontrast images in the depiction of HCC in cirrhotic livers (10,27). Furthermore, diversion of portal venous blood to the spleen in portal hypertension decreases blood flow to the liver; this may decrease the amount of ferumoxides arriving in the liver. Yamashita et al (28) suggested that ferumoxides perform well in all cases of chronic liver disease except in the most severe cases of cirrhosis. Yamamoto et al (29) reported better detection of HCC in cirrhotic livers with ferumoxides than with unenhanced MR imaging.
The dose of ferumoxides may be a factor in the detectability of hepatic lesions in patients with liver cirrhosis. Tang et al (10) and Matsuo et al (26) each administered 10 µmol/kg of ferumoxides and achieved sensitivity values of 78% (80 of 103 lesions) and 62% (150 of 234 segments), respectively. We administered ferumoxides at a dose of 15 µmol/kg, and the results were much better. It may be postulated that the increased amount of iron in circulation may increase the iron uptake by Kupffer cells in diseased liver. Scott et al (30) reported that administration of high doses (15 µmol/kg) and low doses (7.5 µmol/kg) of ferumoxides to patients with colorectal cancer during evaluation for hepatic metastasis did not affect the lesion-to-liver contrast-to-noise ratio; however, the patients in this study did not have cirrhosis.
In the present study, 45 (64%) of 70 patients had mild cirrhosis, 10 (14%) had chronic hepatitis, all patients underwent MR imaging with no undesirable side effects, and all the images obtained were satisfactory. Therefore, the increased total amount of ferumoxides might have had a positive effect in our study, yielding an increased rate of depiction of HCC nodules in patients with mild cirrhosis. Ferumoxides-enhanced MR imaging is limited by the fact that it is difficult to detect and differentiate well-differentiated HCC and dysplastic nodules with this modality (31,32). In our study, there was one case of well-differentiated HCC that was detected at triphasic helical CT but barely detected at ferumoxides-enhanced MR imaging (Fig 4).
This study had several limitations. First, unenhanced MR images were not additionally obtained in this study because of the long time required for biphasic (ie, unenhanced and ferumoxides-enhanced) MR imaging. Unenhanced MR imaging is known to be necessary to decrease the number of false-positive findings attributed to the presence of small vessels. However, the results of this study included only a few false-positive findings. Second, because the selection of candidates for surgical resection was influenced by findings at triple-phase helical CT and ferumoxides-enhanced MR imaging, which themselves were being investigated for their efficiency in the depiction of HCC, strong positive bias was present.
Third, we used a relatively slow rate (3 mL/sec) for contrast material injection during CT examination; a faster injection rate (4–5 mL/sec) is used at many institutions. However, the body weight of our patients is typically lower than that of Western patients, so the injection rate of 3 mL/sec was reasonably rapid for our patients. We used a 7-mm section thickness at CT, which is fairly thick by today’s standards. Detectability of HCC with helical CT may be improved by using multi–detector row CT with thin collimation because of the improved spatial and temporal resolution of this technique. Another limitation of our study was that the observers were aware that all of the patients were found to have HCC at histopathologic examination after hepatic resection surgery.
In summary, the results of our study show that ferumoxides-enhanced MR imaging is superior to triple-phase helical CT for depicting HCC, and we therefore recommend the use of ferumoxides-enhanced MR imaging in the preoperative staging work-up of candidates for surgical resection of HCC.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, B.K.K., J.H.L., D.C.; study concepts, J.H.L., D.C.; study design, J.H.L.; literature research, B.K.K., J.H.L., D.C.; clinical studies, J.H.L., S.H.K., D.C., H.K.L., W.J.L., S.J.L; data acquisition, B.K.K., D.C.; data analysis/interpretation, B.K.K., D.C., J.H.L.; statistical analysis, B.K.K., D.C.; manuscript preparation, B.K.K., D.C., J.H.L.; manuscript definition of intellectual content, J.H.L.; manuscript editing and revision/review, D.C., J.H.L.; manuscript final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. J. Kim, J. M. Lee, J. K. Han, K. H. Kim, J. Y. Lee, and B. I. Choi Peripheral Mass Forming Cholangiocarcinoma in Cirrhotic Liver Am. J. Roentgenol., December 1, 2007; 189(6): 1428 - 1434. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. K. Kim, H. S. Kwak, C. S. Kim, G. H. Chung, Y. M. Han, and J. M. Lee Hepatocellular Carcinoma in Patients with Chronic Liver Disease: Comparison of SPIO-enhanced MR Imaging and 16-Detector Row CT Radiology, December 21, 2005; (2005) 2381042193. [Abstract] [Full Text]
|
|
|
|
|
|
A. M. Lutz, J. K. Willmann, K. Goepfert, B. Marincek, and D. Weishaupt Hepatocellular Carcinoma in Cirrhosis: Enhancement Patterns at Dynamic Gadolinium- and Superparamagnetic Iron Oxide-enhanced T1-weighted MR Imaging Radiology, November 1, 2005; 237(2): 520 - 528. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. H. Kim, D. Choi, S. H. Kim, J. H. Lim, W. J. Lee, M. J. Kim, H. K. Lim, and S. J. Lee Ferucarbotran-Enhanced MRI Versus Triple-Phase MDCT for the Preoperative Detection of Hepatocellular Carcinoma Am. J. Roentgenol., April 1, 2005; 184(4): 1069 - 1076. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
