| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology, Thomas Jefferson University Hospital, 132 S 10th St, 1096 Main Bldg, Philadelphia, PA 19107. Received September 7, 2001; revision requested November 8; revision received December 19; accepted January 29, 2002. Address correspondence to D.G.M. (e-mail: donald.mitchell@mail.tju.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Two hundred thirty-six patients (121 with pathologically proved cirrhosis and 115 without history of chronic hepatic diseases) underwent magnetic resonance (MR) imaging. Two independent observers measured C/RL-r and compared it with C/RL-m. Results were compared by using receiver operating characteristic (ROC) curves and accuracy measures at various thresholds.
RESULTS: The area below the ROC curve was greater for C/RL-r (0.797) than for C/RL-m (0.731; P = .040). By using a C/RL-r greater than 0.90, the sensitivity, specificity, and accuracy for the MR imaging diagnosis of cirrhosis were 71.7%, 77.4%, and 74.2%, respectively. The highest accuracy of the C/RL-m was 65.7%, when the C/RL-m was greater than 0.55. Interobserver agreement was statistically confirmed for both measurements by using 
analysis. Significant differences were found among the three Child-Pugh classes by using C/RL-r (P = .0105) but not by using C/RL-m.
CONCLUSION: C/RL-r is more accurate for diagnosing cirrhosis and evaluating its clinical severity than is C/RL-m.
© RSNA, 2002
Index terms: Liver, cirrhosis, 761.794 • Liver, MR, 761.121411, 761.121412, 761.121415, 761.12143 • Liver, size
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
One imaging sign for diagnosing cirrhosis is the caudate–right lobe ratio (C/RL) that Harbin et al (1) proposed. These authors chose the bifurcation of the main portal vein, a reproducible landmark, to divide these lobes. A ratio greater than 0.65 was considered sensitive and specific for cirrhosis in their series.
One cause of hypertrophy of the caudate lobe and atrophy of the right is thought to be changes in their blood supply. Although most portal branches distributing to the caudate lobe arise from the left portal branch or from the bifurcation of the portal vein, there are branches from the right main portal vein and even the posterior segmental branch (6). We have also observed that the hypertrophied portion of liver appears to extend beyond the bifurcation of the main portal vein, to the right lateral edge of the ligamentum venosum, where the right portal vein bifurcates.
We therefore speculated that the right portal venous bifurcation might more accurately divide the hypertrophied caudate and central liver from the atrophied right lobe than does the main portal venous bifurcation. Thus, the purpose of our study was to determine whether a modified C/RL with use of the right portal vein to set the lateral boundary (C/RL-r) is more accurate for diagnosing cirrhosis and evaluating its clinical severity than is the previously described C/RL with use of the main portal vein to set the lateral boundary (C/RL-m).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The cirrhosis group consisted of 80 men and 41 women (age range, 22–84 years; mean age, 54.0 years). Among the 80 men with cirrhosis, age range was 26–84 years (mean age, 54 years). Among the 41 women with cirrhosis, age range was 29–84 years (mean age, 52.8 years).
The control group consisted of 48 men and 67 women (age range, 22–97 years; mean age, 52.5 years). Among the 48 men in the control group, age range was 26–97 years (mean age, 54.0 years). Among the 67 women in the control group, age range was 22–81 years (mean age, 51.5 years).
Patients in the cirrhosis group had been referred for MR imaging to evaluate the severity of cirrhosis and portal hypertension, to perform preoperative studies before hepatic transplantation, or to screen for or further examine hepatic lesions that were suspected because of results obtained with other imaging modalities. At our institution, MR imaging is performed annually to evaluate patients on the waiting list for hepatic transplantation. Patients in the control group underwent imaging to detect or characterize masses. Reports indicated normal results in 98, including five with no evidence of metastases; cavernous hemangiomas in 10; simple cysts in six; and focal nodular hyperplasia in one.
One of the authors (H.A.) retrospectively reviewed all pathology reports that were entered in the patients’ clinical history. Cirrhosis was caused by hepatitis B (n = 14), hepatitis C (n = 43), hepatitis B and C (n = 3), alcohol abuse (n = 19), alcohol abuse with hepatitis C (n = 9), primary biliary cirrhosis (n = 7), 
1-antitrypsin deficiency (n = 4), Wilson disease (n = 1), cryptogenic origin (n = 10), sarcoidosis (n = 1), or autoimmune hepatitis (n = 2), or cause was undetermined due to insufficient clinical data (n = 8). Cirrhosis was confirmed by means of percutaneous hepatic biopsy in 81 patients, hepatic transplantation in 18, transjugular hepatic biopsy in 19, and autopsy in three. The control group had undergone MR imaging for reasons other than chronic hepatic disease (eg, patients suspected of having hepatic cyst or hemangioma or benign disease of other organs).
We were able to determine Child-Pugh classification (7,8) from available clinical records in 38 of 121 patients with cirrhosis. Nine patients were classified as having Child A, 10 patients as having Child B, and 19 patients as having Child C.
Pathology grading of cirrhosis activity, as indicated in the pathologist’s report, could be determined for 115 patients. It was mild (n = 37), moderate (n = 38), or severe (n = 40). For six patients, activity was not indicated in the biopsy report.
MR Imaging Technique
All MR imaging was performed with 1.5-T units (Signa Magnetom H-15; GE Medical Systems, Milwaukee, Wis) by using transverse T1- and T2-weighted techniques. T1-weighted imaging included one or more of following sequences: conventional spin echo 400–600/11–22 (repetition time msec/echo time msec), in-phase gradient echo (80–210/4.0–4.8) with a 70°–90° flip angle, and opposed-phase gradient echo (80–210/1.6–2.7 or 6.0–7.0) with a 60°–90° flip angle. T2-weighted imaging included the following sequences: conventional spin echo (1,500–3,000/50–100), breathing-averaged fast spin echo (3,000–7,500/91–104 [effective]) with or without fat suppression, and breath-hold fast spin echo (2,500–4,200/70–138 [effective]).
All but seven of the 236 patients also underwent dynamic multiphasic imaging that consisted of T1-weighted spoiled gradient-echo imaging after intravenous injection of 0.1 mmol of gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) per kilogram of body weight. The dynamic technique consisted of either two-dimensional multisection (80–210/1.6–2.7 or 6.0–7.0) with a 60°–90° flip angle or three-dimensional (7/2.2) with a 15° flip angle. All dynamic imaging was followed by two-dimensional fat-suppressed imaging (80–210/1.8–2.3) with a 70°–90° flip angle. Other imaging parameters included a 256 x 128–256 imaging matrix, usually with use of a rectangular field of view to reduce the number of phase-encoding views, and sections 7–12-mm thick with section gap of 2 mm or less.
Hepatic Measurement
Images of subjects in the control and cirrhosis groups were presented to reviewers in random order. Measurements were obtained on a clinical workstation (Canon Medical Systems, Irvine, Calif), when possible (134 patients). When electronic access to images was not available (102 patients), measurements were obtained from hard-copy images. Images were measured without access to clinical data, independently and separately, by two radiologists (H.A., T.K.) who had several years of experience reading abdominal MR images, were familiar with the diagnosis of hepatic diseases, and had not previously viewed the cases.
Each reviewer individually chose images for measurement on the basis of clarity of the landmarks, which were the bifurcation of the main portal and right portal veins. In most cases, the portal phase images obtained after administration of contrast material were chosen for measurement. If the observer considered that the landmarks were not best on these images, images from other sequences were used.
C/RL-m.—As described by Harbin et al (1), and illustrated in Figure 1, a line (line 1) was drawn parallel to the midsagittal plane through the right lateral wall of the bifurcation of the main portal vein. A second line (line 2) was drawn parallel to line 1 through the most medial margin of the caudate lobe. A third line (line 3) was drawn perpendicular to lines 1 and 2 midway between the main portal vein and the inferior vena cava and extended to the right lateral margin. The distances along line 3 between the lines 1 and 2 (Fig 1, C) and along line 3 between the right lateral margin and line 1 (Fig 1, R) were measured and expressed as the ratio of C to R, which was termed the C/RL-m.
|
Statistical Analysis
Interobserver variability for each measurement was assessed by using the coefficient of variation (SD/mean x 100) (9) and value at each threshold of C/RL-m and C/RL-r. Average values of both observers’ measurements were used for further data analysis. C/RL-m and C/RL-r were compared for the cirrhosis group versus the control group by using the unpaired two-tailed Student t test.
Receiver operating characteristic (ROC) curves were constructed to compare the C/RL-r with the C/RL-m (10). True-positive cases were defined as cirrhosis cases that were correctly assigned. False-positive cases were defined as control cases that were incorrectly assigned. The diagnostic capability was determined by calculating the area below the ROC curve (Az) for each ratio-specific result. The Az values for C/RL-m and for C/RL-r were then compared by using the bivariate 
2 test.
The sensitivity, specificity, accuracy, and positive predictive value of the C/RL-m and the C/RL-r for the MR imaging diagnosis of cirrhosis were calculated at values of 0.05 increments for each ratio. The C/RL-r in the cirrhosis group was compared with that in the control group by means of 2 analysis. A P value of less than .05 was considered to indicate a statistically significant difference. All data were statistically analyzed with the Student t test and the 2 test.
We compared the two ratios (C/RL-m and C/RL-r) for the different causes of cirrhosis. Causes that occurred in only one or two patients (Wilson disease, sarcoidosis, autoimmune hepatitis) were considered, along with those of undetermined cause or insufficient clinical data, as miscellaneous causes. The resultant nine groups (alcohol abuse, alcohol abuse with hepatitis C, 1-antitrypsin deficiency disease, hepatitis B, hepatitis B and C, hepatitis C, cryptogenic origin, primary biliary cirrhosis, and miscellaneous causes) were compared for the two ratios by using the Kruskal-Wallis analysis of variance. The results of C/RL-m and C/RL-r were also compared with regard to clinical severity (Child A, Child B, and Child C) and histologic activity (mild, moderate, and severe), again by using the Kruskal-Wallis analysis of variance, with the Scheffé test used for comparing the two groups with each other.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The means of the C/RL-m values were 0.433 ± 0.112 for the control livers and 0.542 ± 0.143 for the cirrhotic livers (P < .001). The means of the C/RL-r values were 0.771 ± 0.216 for the control livers and 1.036 ± 0.249 for the cirrhotic livers (P < .001).
The ROC curves for diagnosing cirrhosis are shown in Figure 2. Az values were significantly different: 0.737 for C/RL-m and 0.791 for C/RL-r (P = .040).
|
values between two observers were higher for a C/RL-r of greater than 0.90 than for a C/RL-m of greater than 0.55 (0.7195 vs 0.628), which suggests greater reproducibility for C/RL-r. Interobserver agreements were statistically comfirmed in both measurements (P < .001 for both).
|
Clinical Correlation
Significant differences were found between C/RL-m (P = .032) and C/RL-r (P = .012) among the nine cause groups. Figure 3 shows the mean of each ratio plus or minus SD according to diagnosis. The mean for C/RL-r (1.197 ± 0.172) was highest for hepatitis B–related cirrhosis. The mean for C/RL-m (0.633 ± 0.098) was highest for primary biliary cirrhosis.
|
Figures 4–6 illustrate examples of the appearance of high C/RL-r and normal C/RL-m in patients with different levels of clinical severity of cirrhosis.
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The causes of hypertrophy of the caudate lobe and atrophy of the right lobe and the medial segment are unclear but are thought to be linked to alterations in portal blood flow, since segmental hepatic volume is related to portal venous blood flow because of various trophic factors within portal venous blood (11). Hepatic fibrosis causes attenuation of the intrahepatic portal and hepatic venous branches, and the hepatic vascular bed is reduced (12). Impaired drainage of blood from the liver, caused by compression of hepatic venous tributaries by regenerating nodules or fibrosis, increases the resistance to portal flow (13).
A predominance of blood supply to the caudate lobe arises from the left branches of the portal vein or from the bifurcation of the portal vein (14); however, a branch to the caudate process from the posterior segmental branch is present in 51% (6). The right main portal venous branch consists of a short trunk of greater diameter than that of the left main branch, which divides after a short distance into several secondary branches (15). The caudate branches have a shorter intrahepatic course than do other vessels, possibly being a factor in relative preservation of caudate lobe blood supply (15).
We suspect that proximity of parenchymal tissue to the main portal venous branches and/or access to preserved caudate hepatic venous drainage may protect this region of the liver from atrophy, but we were not able to evaluate the specific causes of atrophy and hypertrophy. Rather, we evaluated an alternative ratio, C/RL-r, by using the right portal venous bifurcation to divide the hypertrophied and atrophied regions of liver. This landmark appears to correspond more closely to the hypertrophied and atrophied portions of liver. Comparing these two ratios, we found a higher Az value for C/RL-r (0.797) than for Harbin’s ratio, C/RL-m (0.737; P = .040). The interobserver agreement obtained by using the value was statistically confirmed for both C/RL-r and C/RL-m. This result indicates that the landmark of right portal bifurcation is also easy to reproduce between observers.
In our study, we had Child-Pugh staging for only about one-third of the patients, half of whom were in the Child C stage. We are therefore reluctant to make any firm conclusions regarding Child-Pugh staging on the basis of our data. Although one might expect that the C/RL would be highest with the most severe cirrhosis, that is, Child C, we found that C/RL-r was highest for Child B in our series. Although we are not certain of the explanation for this, it is possible that the mean ratio for Child C may have been reduced by inclusion of patients with advanced atrophy of the whole liver; in these livers, the caudate lobe might actually have been partially atrophied. This explanation remains conjectural, however, and must be verified by replication in a larger series with more controlled patient selection.
Several investigators (1,3–5,16–18) have reported morphologic change of the caudate lobe for diagnosing cirrhosis. Giorgio et al (16) reported a high sensitivity of the C/RL with ultrasonography (US) for diagnosing hepatitis B virus–related cirrhosis. We found that with C/RL-r, the mean of hepatitis B virus–related cirrhosis was the highest, but with C/RL-m the mean of primary biliary cirrhosis was the highest. Dodd et al (17) reported that among patients with end-stage hepatic disease, caudate hypertrophy was most prominent in patients with primary sclerosing cholangitis; the population in our study did not include patients with primary sclerosing cholangitis and was not restricted to patients with end-stage disease.
More recently, Okazaki et al (18) noted that caudate volume was greater in cirrhosis secondary to alcohol abuse than it was in virally induced cirrhosis. Our linear measurements of C/RL-m and C/RL-r did not show a significant difference between these populations. Similarly, Giorgio et al (16) reported that the sensitivity of the C/RL at US was low in alcoholic cirrhosis. Thus, the difference of cause and clinical and pathologic severity alter the sensitivity, specificity, and accuracy of diagnosing hepatic cirrhosis obtained by measuring the atrophied right lobe and hypertrophed caudate lobe.
Although Harbin et al (1) did not state the cause of cirrhosis in their series, we suspect that alcohol abuse accounted for a higher percentage of cirrhosis cases in their population, which was drawn from a North American practice more than 20 years ago. At that time, alcohol abuse accounted for most cirrhosis in the United States (19,20). Harbin et al did not indicate the clinical or pathologic severity of cirrhosis; different selection criteria probably account for the much lower accuracy of C/RL-m in our study than in theirs.
Our observations comparing two different C/RLs can probably apply to computed tomography (CT), as well as to MR imaging, since the same landmarks are visible. We evaluated MR imaging rather than CT in our series because at our institution MR imaging is preferred for routine examination of most patients with hepatitis and cirrhosis.
Our study is limited by how the subjects were selected. The control patients were chosen randomly on the basis of the use of imaging technique and period of imaging identical to that of the patients with cirrhosis. We did not attempt to match sex, race, or socioeconomic status between groups. However, we are not aware of any data to suggest that hepatic morphology differs with respect to these variables.
Since the control population was drawn from among people examined with MR imaging performed because they were suspected of having disease, this population may not be representative of the general population or of patients without cirrhosis who are at risk for cirrhosis. Additionally, patient selection criteria in our study, which were based on MR imaging records and/or clinical requests, may have been biased because patients with cirrhosis who did not have clinically or radiologically abnormal findings might have been overlooked. This limitation is a common one in retrospective studies. The accuracy of the C/RL could be more accurately estimated if patients were selected completely independently of whether the MR images showed evidence of cirrhosis. A more clinically relevant assessment of the usefulness of the C/RL would involve comparison between patients with cirrhosis and patients without cirrhosis who are at risk for cirrhosis.
These limitations prevent us from stating or generalizing about the accuracy of either of these C/RLs to other populations. However, all of these weaknesses apply to both of the C/RLs, and we found that C/RL-r was more sensitive. Even though we found C/RL-r more sensitive than C/RL-m, it is not sufficiently sensitive to exclude early cirrhosis, and other imaging and clinical data are needed. However, abnormal C/RL-r results can help by being suggestive of cirrhosis.
In conclusion, the C/RL-r appears to better represent the division between hypertrophied and atrophied liver and is more sensitive and accurate for diagnosing cirrhosis than is the previously used version of the C/RL.
|
FOOTNOTES |
|---|
3 Current address: Department of Radiology, Hokkaido University School of Medicine, Sapporo, Japan.
Abbreviations: Az = area below the ROC curve, C/RL = caudate–right lobe ratio, C/RL-m = C/RL with use of the main portal vein to set the lateral boundary, C/RL-r = C/RL with use of the right portal vein to set the lateral boundary, ROC = receiver operating characteristic
Author contributions: Guarantors of integrity of entire study, H.A., D.G.M.; study concepts and design, H.A., D.G.M.; literature research, H.A., D.G.M.; clinical studies, H.A., D.G.M., T.K.; data acquisition, H.A., D.G.M., T.K.; data analysis/interpretation, all authors; statistical analysis, H.A., D.G.M.; manuscript preparation, H.A., D.G.M.; manuscript definition of intellectual content, H.A., D.G.M., G.H., K.I.; manuscript editing, revision/review, and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. C. Tan The Right Posterior Hepatic Notch Sign Radiology, July 1, 2008; 248(1): 317 - 318. [Full Text] [PDF]
|
|
|
|
 |
|
 A. S. d. A. Bezerra, G. D'Ippolito, R. P. Caldana, D. D. Leopoldino, G. R. Batista, D. R. Borges, G. de Jesus Lopes Filho, and M. Ahmed Differentiating Cirrhosis and Chronic Hepatosplenic Schistosomiasis Using MRI Am. J. Roentgenol., March 1, 2008; 190(3): W201 - W207. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Hagiwara, H. Rusinek, V. S. Lee, M. Losada, M. A. Bannan, G. A. Krinsky, and B. Taouli Advanced Liver Fibrosis: Diagnosis with 3D Whole-Liver Perfusion MR Imaging--Initial Experience Radiology, March 1, 2008; 246(3): 926 - 934. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Atassi, A. K. Bangash, A. Bahrani, G. Pizzi, R. J. Lewandowski, R. K. Ryu, K. T. Sato, V. L. Gates, M. F. Mulcahy, L. Kulik, et al. Multimodality Imaging Following 90Y Radioembolization: A Comprehensive Review and Pictorial Essay RadioGraphics, January 1, 2008; 28(1): 81 - 99. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Taouli, A. J. Tolia, M. Losada, J. S. Babb, E. S. Chan, M. A. Bannan, and H. Tobias Diffusion-Weighted MRI for Quantification of Liver Fibrosis: Preliminary Experience Am. J. Roentgenol., October 1, 2007; 189(4): 799 - 806. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
X.-p. Zhou, T. Lu, Y.-g. Wei, and X.-z. Chen Liver Volume Variation in Patients with Virus-Induced Cirrhosis: Findings on MDCT Am. J. Roentgenol., September 1, 2007; 189(3): W153 - W159. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Kato, M. Kanematsu, X. Zhang, M. Saio, H. Kondo, S. Goshima, and H. Fujita Computer-Aided Diagnosis of Hepatic Fibrosis: Preliminary Evaluation of MRI Texture Analysis Using the Finite Difference Method and an Artificial Neural Network Am. J. Roentgenol., July 1, 2007; 189(1): 117 - 122. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Yoshikawa, H. Kawamitsu, D. G. Mitchell, Y. Ohno, Y. Ku, Y. Seo, M. Fujii, and K. Sugimura ADC measurement of abdominal organs and lesions using parallel imaging technique. Am. J. Roentgenol., December 1, 2006; 187(6): 1521 - 1530. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Vilgrain, B. Condat, C. Bureau, A. Hakime, A. Plessier, D. Cazals-Hatem, and D. C. Valla Atrophy-Hypertrophy Complex in Patients with Cavernous Transformation of the Portal Vein: CT Evaluation Radiology, October 1, 2006; 241(1): 149 - 155. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. A. Aguirre, C. A. Behling, E. Alpert, T. I. Hassanein, and C. B. Sirlin Liver Fibrosis: Noninvasive Diagnosis with Double Contrast Material-enhanced MR Imaging. Radiology, May 1, 2006; 239(2): 425 - 437. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
