| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Russell H. Morgan Department of Radiology and Radiological Sciences (H.J.V.B., D.A.B.) and Department of Surgery and Oncology (M.A.C.), Johns Hopkins University School of Medicine, 600 N Wolfe St, Baltimore, MD 21287; Department of Radiology-MRI, New York University Medical Center, New York City (V.S.L.); Department of Radiology, Duke University Medical Center, Durham, NC (E.K.P.); and Department of Radiology, University of Pennsylvania Medical Center, Philadelphia (E.S.S.). Received May 10, 2001; revision requested June 25; revision received August 17; accepted October 8. Supported by an unrestricted grant from Nycomed-Amersham. Address correspondence to D.A.B. (e-mail: dbluemke@jhmi.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-five patients known to have or suspected of having hepatic lesions who were eligible for surgery underwent dual-phase (ie, arterial and portal phase) helical CT and phased-array MR imaging (ie, unenhanced fast spin-echo T2-weighted imaging and gradient-echo T1-weighted imaging performed before and after administration of mangafodipir trisodium). All images were reviewed independently by three off-site blinded reviewers who separately reviewed the CT scans and MR images. The standard of reference was findings at surgery, intraoperative ultrasonography (US), and histopathologic examination. ROC curves were established to analyze the results for each reader and modality.
RESULTS: Ninety-four lesions (77 malignant and 17 benign) were revealed at surgery, intraoperative US, and/or histopathologic examination. The overall rate of lesion detection for the three readers at CT was 81.9% ± 7.8, 90.4% ± 5.9, and 76.6% ± 8.6. At MR imaging, the detection rates were 72.3% ± 9.0, 71.3% ± 9.1, and 69.1% ± 9.3 (P = .001 for the difference between MR and CT). The average rate of false-positive diagnoses in patients was 14.1% at CT and 6.4% at MR imaging (P = .06 for the difference between MR and CT). The mean areas under the alternative-free-response ROC curves were 0.74 for MR and 0.72 for CT (P = .751, not significant).
CONCLUSION: In detection and characterization of liver lesions, manganese-enhanced MR imaging and dual-phase helical CT were not statistically different.
© RSNA, 2002
Index terms: Liver neoplasms, CT, 761.12114 • Liver neoplasms, MR, 761.12143 • Magnetic resonance (MR), contrast media • Manganese
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Hepatic-specific contrast agents for magnetic resonance (MR) imaging have recently been developed and have the potential to improve the detection of focal hepatic lesions (7,8). Mangafodipir trisodium is a hepatocyte-specific paramagnetic contrast agent that causes predominantly T1 shortening. After intravenous administration of mangafodipir trisodium, the normal liver shows increased signal intensity on T1-weighted images, while metastatic lesions are unchanged in signal intensity. A multicenter phase III trial of mangafodipir trisodium reported that detection, classification, and diagnosis of focal liver lesions at manganese-enhanced MR imaging was comparable or superior to that attained at unenhanced MR imaging and at enhanced CT (9). Bartolozzi et al (10) reported that MR imaging with the use of mangafodipir trisodium is significantly more sensitive than unenhanced MR imaging and is similar in sensitivity to dual-phase helical CT for detection of hepatocellular carcinoma in cirrhosis. However, little is known about the performance of manganese-enhanced MR imaging compared with that of dual-phase helical CT for lesion detection preoperatively.
The purpose of this study was to compare, by means of receiver operating characteristic (ROC) analysis, dual-phase helical CT and manganese-enhanced MR imaging in the detection and characterization of hepatic lesions in patients prior to planned hepatic surgery or intervention (ie, ablation).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MR Imaging
All MR imaging was performed with 1.5-T MR imagers (Signa LX; GE Medical Systems, Milwaukee, Wis). Phased-array surface coils were used in all patients for signal reception. Before injection of the contrast agent, fast spin-echo T2-weighted images (repetition time msec/echo time msec, 3,500–6,000/100) were obtained with an echo train length of 16, four signals average, a matrix of 256 x 256, chemical shift fat suppression with manual tuning of the fat-suppression pulse, and 7-mm section thickness with zero gap (interleaved acquisition). In addition, T1-weighted fast multiplanar spoiled gradient-echo images (110/4.4, 70° flip angle) were obtained with one signal average, a matrix of 256 x 192, 7-mm section thickness, and zero gap (interleaved). Six T1-weighted images were obtained during a breath hold; three to five breath holds were required to image the entire liver (depending on liver size).
Both T1- and T2-weighted images were obtained in the transverse plane. The field of view was 32–40 cm and was adjusted for patient size. A three-quarter field of view was used in the phase-encoding direction. After the unenhanced images were obtained, 5 µmol/kg of body weight (0.1 mL/kg; maximum dose given, 
15 mL) of mangafodipir trisodium (Teslascan; Nycomed, Princeton, NJ) was infused intravenously over 3 minutes and followed by a 10-mL flush of normal saline. The same T1-weighted fast multiplanar spoiled gradient-echo sequence was then repeated 15 minutes after intravenous administration of mangafodipir trisodium.
CT Scanning
Dual-phase helical CT imaging was performed with a Somatom Plus 4 scanner (Siemens Medical Systems, Iselin, NJ). Although most patients were anticipated to have hypovascular metastases (eg, colorectal metastases), dual-phase scanning was performed as the routine for all patients before surgery to aid in lesion characterization and to help identify hypervascular lesions (4–6,10). Scanning parameters with the Somatom scanner were 120 kV, 210 mA, 5-mm collimation, table speed of 5–7.5 mm/sec (pitch of 1–1.5, adjusted on the basis of patient size and breath-hold capacity), 5-mm reconstruction interval, and 1-second scanning time. The arterial phase of scanning started 20 seconds after the start of injection of 150 mL of a nonionic contrast material (300 mg of iodine per milliliter, iohexol, Omnipaque; Nycomed) at 4 mL/sec. The portal phase of scanning began 70 seconds after the start of contrast material injection. CT images were filmed with the standard window and width settings used at our institution for viewing the liver parenchyma.
Standard of Reference
The combination of intraoperative ultrasonography (US) and careful surgical palpation and inspection of the liver, together with histopathologic findings, constituted the standard of reference. All surgery was performed by one experienced hepatobiliary surgeon (M.A.C.) who was aware of the preoperative imaging findings. Before surgery, imaging findings at CT and MR were jointly reviewed by the radiologist (D.A.B.) and the surgeon so that a one-to-one correlation between preoperative and surgical findings could be performed. At the time of surgery, all patients underwent intraoperative US performed with 4–8-MHz dedicated intraoperative sonography probes (ATL 5500; ATL Ultrasound, Bothell, Wash) and lesion biopsy. The number of lesions was recorded according to standard templates that also indicated the size and segmental location of the lesions. In resected segments, specimens were sliced at 1-cm intervals and were examined histopathologically to determine the final diagnosis.
CT and MR imaging examinations were performed within 1 day of each other in 18 patients and within 1 week of each other in seven patients. The average time between the CT and the MR imaging examinations was 2 days (range, 1–7 days). The average time between the initial imaging study (CT or MR imaging) and the surgery was 10 days (minimum, 2 days; maximum, 43 days).
Image Analysis
Three observers (V.S.L., E.S.S., E.K.P.) who were unaware of patient identification and clinical history and who were at institutions other than where the MR imaging, CT, and surgical procedures were performed participated in the image analysis. All observers were experienced in interpreting both CT and MR images, with 4, 8, and 9 years of experience. The MR and CT images were independently reviewed by each observer and were interpreted at discrete sessions separated by an average of 3 weeks. All MR images (ie, those obtained both before and after mangafodipir trisodium enhancement) were combined to form the results of the MR examination. Any identifying information on the CT and MR images was masked. The images were shown in random order to the blinded observers. A standardized template form for each examination was completed on which the interpreter indicated the segmental location of each lesion. If a lesion crossed segmental boundaries, the lesion was assigned to the segment with the greatest involvement.
Each observer recorded the number of hepatic lesions seen, their size, the segmental location of lesions according to the classification scheme of Couinaud, and whether the lesions appeared to be benign or malignant. A rating of diagnostic confidence in the overall interpretation was also assigned for lesion classification, where 1 is very low, 2 is low, 3 is moderate, 4 is high, and 5 is very high. If more than 10 benign or malignant lesions were present, the observer indicated that more than 10 benign or malignant lesions were present and did not add further comments to subclassify confidence levels for individual lesions.
Two radiologists (D.A.B., H.J.V.B.) evaluated the readers’ interpretations in consensus to determine reasons for false-positive or false-negative diagnoses. All imaging studies, together with histopathologic and surgical reports, were available for this review.
Statistical Analysis
Alternative-free-response ROC curves (11) were calculated for each observer by plotting the true-positive fraction against the likelihood of obtaining a false-positive image (ie, an image with one or more false-positive lesions). The conventional ROC method does not allow the recording or differentiation of multiple responses per image, whereas alternative-free-response ROC is a modified ROC technique that allows multiple responses, enabling all of the observers’ responses to be correlated with the actual lesions present (12). As with the conventional ROC method, the area under each alternative-free-response ROC curve was taken to indicate the overall performance of both the modalities and the observers. Sensitivity was defined as the number of true-positive diagnoses at a confidence level of 4 or 5. A false-positive diagnosis was defined as when one or more lesions were identified in a patient by a reader at a confidence level of 4 or 5 when no lesion was present. Specificity was defined as 1 minus the false-positive fraction, where the false-positive fraction was equal to the number of false-positive diagnoses divided by the total number of patients (n = 25).
To estimate the variance in the area under the ROC curve, a bootstrap analysis was performed by using 1,000 random samples obtained by sampling the results in the 25 patients with replacement. Because the number of lesions varied by patient, each sample had a different total number of true lesions. ROC curves were created from the data for all combinations of reader and imaging modality. Differences in the area under the ROC curve between methods were calculated by averaging the difference in area-under-the-curve values over the three readers. These averages gave an estimate of the variance in the area under the curve difference between the MR and the CT imaging techniques. The probability of the difference between the MR and CT values being not equal to zero was determined by transforming the data to a standard normal distribution with the statistic Z = mean/SD. The agreement between readers in classifying each lesion as benign or malignant was assessed with the McNemar test, with a two-tailed P < .05 defined as statistically significant.
The absolute detection rate was defined as the number of lesions identified by the reader divided by the total number of actual lesions (irrespective of their classification as benign or malignant). The false-positive rate was defined as the number of patients with at least one false-positive diagnosis (ie, a lesion identified as present when it was not present at histopathologic examination) divided by the total number of patients (n = 25). Significance testing was performed by using 
2 analysis with Statview 5.0 (SAS Institute, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
Of the 94 lesions, six (6%) in three patients were not detected preoperatively at CT or at MR imaging by any of the readers. Five of these lesions were metastatic lesions and one was a granuloma. These lesions were all small—four lesions were less than or equal to 3 mm in diameter, and two lesions were 5–10 mm in diameter. The lesions less than or equal to 3 mm in diameter were detected at histopathologic examination of a resected liver segment and were not seen at intraoperative US. For paired lesion evaluation, there was no statistically significant difference in the classification of lesions as benign or malignant at MR image interpretation between any of the groups (P > .05). At CT scan interpretation, differences in classification were significant between reader 1 and reader 2 (P < .002), but not for other reader pairs.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
When we used strict criteria that required readers to classify lesions as definitely benign or malignant, the sensitivity of both CT and MR was considerably lower than the overall lesion detection rate (Tables 1, 2). We believe that this is due in part to variation between readers for confidence in lesion diagnosis, as well as to the use of liver-segment scoring to define the presence or absence of a lesion (6). However, the specificity values were similar to previously reported values in the literature (6,7,13–15).
Eighty-eight (94%) of 94 lesions were detected at either MR imaging or CT; six lesions (6%) could not be detected on preoperative images even after the images were reviewed in comparison with surgical and histopathologic findings. Four of these six lesions measured less than 5 mm in diameter; two measured 5–10 mm in diameter. All lesions of 10 mm or less could be detected at CT and/or MR imaging. All four lesions smaller than 5 mm were missed at intraoperative US.
The ideal modality for imaging patients before planned liver surgery should have a high sensitivity but a low false-positive rate. Although CT during arterial portography has been shown to be sensitive for lesion detection (13), its invasiveness, occasional technical failure, and low specificity (ie, high false-positive rate) have reduced its use. In our institution we have replaced the use of CT during arterial portography with the use of dual-phase helical CT and MR imaging. The use of MR imaging is reserved for preoperative staging of lesions that appear to be resectable on the basis of findings at CT and for characterizing the lesions that are equivocal at CT. Intraoperative US is an important aid during hepatic surgery because it can reveal additional small liver tumors that may not be depicted preoperatively even with the use of modern noninvasive imaging modalities (16). However, the sensitivity of intraoperative US is reported to be between 80% and 96% (17,18).
There were several limitations to this study. First, only findings at intraoperative US and liver palpation could be used as the standard of reference for nonresected portions of the liver. Histologic proof was not available for nonresected segments. It is therefore likely that some small lesions were present that were not detected either at preoperative imaging or at the time of surgery. Also, the blinded observers were not asked to evaluate the additive value of each examination. For example, the additive benefit of enhanced MR imaging compared with unenhanced MR imaging versus arterial and portal phase CT images was not specifically addressed. Also, small benign lesions may have been more accurately detected or characterized with the T2-weighted sequence than with the T1-weighted manganese-enhanced sequence. Although similar to other patient sample sizes in the literature, the number of patients in the study was small. Finally, the section thickness used in MR imaging was 7 mm, while the section thickness used in CT was 5 mm. The use of the larger section thickness in MR imaging was based on considerations of adequate signal-to-noise ratio.
In conclusion, detection rates of liver lesions were higher at dual-phase helical CT than at manganese-enhanced MR imaging. There was no statistically significant difference in the area under the alternative-free-response ROC curve for manganese-enhanced MR imaging compared with that of contrast-enhanced dual-phase helical CT in the detection and characterization of focal hepatic lesions.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, D.A.B.; study concepts and design, D.A.B., H.J.V.B.; literature research, H.J.V.B.; clinical studies, E.S.S., E.K.P., V.S.L., M.A.C.; data acquisition and analysis/interpretation, H.J.V.B., D.A.B.; statistical analysis, H.J.V.B., D.A.B.; manuscript preparation, definition of intellectual content, and editing, H.J.V.B., D.A.B.; manuscript revision/review, all authors; manuscript final version approval, H.J.V.B., D.A.B.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. J. Kim, K. W. Kim, J. H. Byun, H. J. Won, Y. M. Shin, P. N. Kim, M.-S. Lee, and M.-G. Lee Comparison of mangafodipir trisodium- and ferucarbotran-enhanced MRI for detection and characterization of hepatic metastases in colorectal cancer patients. Am. J. Roentgenol., April 1, 2006; 186(4): 1059 - 1066. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. A. Bluemke, D. Sahani, M. Amendola, T. Balzer, J. Breuer, J. J. Brown, D. D. Casalino, P. L. Davis, I. R. Francis, G. Krinsky, et al. Efficacy and Safety of MR Imaging with Liver-specific Contrast Agent: U.S. Multicenter Phase III Study Radiology, October 1, 2005; 237(1): 89 - 98. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. M. Fayad, I. R. Kamel, D. G. Mitchell, and D. A. Bluemke Functional MR Cholangiography: Diagnosis of Functional Abnormalities of the Gallbladder and Biliary Tree Am. J. Roentgenol., May 1, 2005; 184(5): 1563 - 1571. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. W. Kim, A. Y. Kim, T. K. Kim, S. H. Park, H. J. Kim, Y. K. Lee, M.-S. Park, H. K. Ha, P. N. Kim, J. C. Kim, et al. Small (l.e. 2cm) Hepatic Lesions in Colorectal Cancer Patients: Detection and Characterization on Mangafodipir Trisodium-Enhanced MRI Am. J. Roentgenol., May 1, 2004; 182(5): 1233 - 1240. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Schoder, S. M. Larson, and H. W.D. Yeung PET/CT in Oncology: Integration into Clinical Management of Lymphoma, Melanoma, and Gastrointestinal Malignancies J. Nucl. Med., January 1, 2004; 45(90010): 72S - 81. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Kettenbach, W. Kostler, E. Rucklinger, B. Gustorff, M. Hupfl, F. Wolf, K. Peer, M. Weigner, J. Lammer, W. Muller, et al. Percutaneous Saline-Enhanced Radiofrequency Ablation of Unresectable Hepatic Tumors: Initial Experience in 26 Patients Am. J. Roentgenol., June 1, 2003; 180(6): 1537 - 1545. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
