| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Departments of Radiology (A.Q., J.S.G., B.M.Y., F.V.C.), Pathology (S.K.), and Medicine (R.B.M.), University of California San Francisco, Box 0628, Room M-372, 505 Parnassus Ave, San Francisco, CA 94143-0628. Received March 23, 2004; revision requested June 2; revision received November 24; accepted January 12, 2005. Address correspondence to A.Q. (e-mail: Aliya.Qayyum{at}radiology.ucsf.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Committee on Human Research approval was obtained. Patient consent was not required. Data collection ended before HIPAA regulations were implemented, but patient anonymity was maintained. Twenty-seven patients, 16 with cirrhosis, were retrospectively identified who underwent MR imaging before histopathologic evaluation of liver fat at biopsy or surgery. The patient population consisted of 15 male and 12 female patients (mean age, 55 years; range, 16–75 years). One radiologist blinded to the histopathologic results recorded mean signal intensity derived from three regions of interest placed in the right and left lobes of the liver on three sections and signal intensity of the spleen from one region of interest within the same section. Liver fat was quantified with the relative loss of signal intensity on out-of-phase images compared with that on in-phase T1-weighted gradient-echo images and with relative loss of signal intensity on T2-weighted fast spin-echo MR images obtained with fat saturation compared with those obtained without fat saturation. Hotelling t test was used to compare correlation coefficients between relative signal intensity differences and histopathologically determined percentage of fat.
RESULTS: In patients without cirrhosis, liver fat quantification with fat-saturated fast spin-echo MR imaging was significantly better than it was with out-of-phase gradient-echo MR imaging (r = 0.92 vs 0.69, P < .01). In patients with cirrhosis, liver fat quantification was correlated only with fat-saturated fast spin-echo MR imaging (r = 0.76, P < .01); the relative signal intensity loss on out-of-phase gradient-echo MR images was not correlated with histopathologically determined percentage of fat (r = 0.25, P = .36).
CONCLUSION: Preliminary results suggest liver fat may be more accurately quantified with fat-saturated fast spin-echo MR imaging than with out-of-phase gradient-echo MR imaging, especially in patients with cirrhosis.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Researchers (5,6) in early studies of liver fat quantification with magnetic resonance (MR) imaging used older methods, such as the modified Dixon technique and fast gradient-echo sequences, and demonstrated a substantial correlation between imaging findings and the histopathologically determined percentage of fat for moderate steatosis. Results for fat quantification with conventional spin-echo T2-weighted MR imaging sequences without fat-saturation have been mixed (5,7), and, to our knowledge, there are no studies in which fat-saturated T2-weighted fast spin-echo MR imaging sequences have been used for this purpose. We are unaware of any studies in which MR imaging evaluation of steatosis in patients with cirrhosis was compared with that in patients without cirrhosis, even though cirrhosis might be a potentially confounding factor. Therefore, we undertook this study to retrospectively determine the relative accuracy of liver fat quantification with out-of-phase gradient-echo MR imaging and fat-saturated fast spin-echo MR imaging in patients with and without cirrhosis, with histologic analysis as the reference standard.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Patients were excluded if (a) an artifact appeared on MR images and precluded accurate measurement of signal intensity (n = 5), (b) slides of pathologic slices were not available (n = 4), and (c) MR images were not retrievable (n = 3).
The final population of 27 patients formed the study group, and it included 15 male and 12 female patients with a mean age of 55 years (range, 16–75 years). The clinical indications for MR imaging in these patients were as follows: cirrhosis of unknown cause in seven patients; chronic hepatitis C with cirrhosis in six patients; chronic hepatitis C without cirrhosis in five patients; hepatic metastases in two patients; chronic Budd-Chiari syndrome with cirrhosis in two patients; cholangiocarcinoma in two patients; and cystic fibrosis with cirrhosis, focal nodular hyperplasia, and biliary cystadenoma in one patient each. A total of 16 of the 27 patients had cirrhosis. There was no difference between the mean age of the patients with cirrhosis and that of the patients without cirrhosis. None of the patients were known to have primary or secondary hemochromatosis.
MR Imaging Technique
MR imaging was performed with a 1.5-T superconducting magnet (Signa; GE Medical Systems, Milwaukee, Wis) by using a four-element torso phased-array coil (GE Medical Systems). Imaging studies were as follows: (a) Transverse T1-weighted in-phase and out-of-phase breath-hold spoiled gradient-echo MR imaging was performed with repetition time msec/echo time msec, 90–200/1.8–2.1 (out-of-phase), 4.2 (in-phase); flip angle, 70°–90°; section thickness, 8 mm; intersection gap, 1 mm; matrix, 256 x 128–192; field of view, 32–40 cm; and signal acquired, one. In 20 patients, in-phase and out-of-phase images were acquired separately. In seven patients, dual-echo acquisition was used. (b) Transverse (n = 21) or coronal (n = 6) T2-weighted breath-hold single-shot fast spin-echo MR imaging without fat saturation was performed with the following parameters: 
/100 (effective); section thickness, 6 mm; intersection gap, 1 mm; matrix, 256 x 160–192; field of view, 32–40 cm; and signal acquired, one. (c) Transverse fat-saturated T2-weighted breathing-averaged fast spin-echo MR imaging was performed with the following parameters: 4000–5000/100 (effective); section thickness, 8 mm; intersection gap, 1 mm; matrix, 256 x 256; field of view, 32–40 cm; and signals acquired, two to three. Fat saturation was applied by using manual frequency selection.
Image Interpretation
One attending radiologist (J.S.G.), who had 6 years of experience in interpretation of liver MR images and was unaware of histopathologic analysis results, reviewed the MR images on a picture archiving and communication system workstation (Impax; Agfa, Mortsel, Belgium). The signal intensity values of regions of interest in the liver and spleen were recorded for in-phase T1-weighted MR images, out-of-phase T1-weighted MR images, T2-weighted MR images without fat saturation, and T2-weighted MR images with fat saturation. These regions of interest were drawn so that they were 1–2 cm in diameter. Three regions of interest were obtained in the liver (two in the right lobe and one in the left lobe) in three sections. The three sections were selected to include levels above, at, and below the portal vein. The standard deviation of the signal intensity measurements within each region of interest was kept to less than 10%, the region of interest included areas of parenchyma that did not contain vessels or artifact, and the region of interest was at a similar location and depth from the liver surface on the paired T1-weighted and T2-weighted MR images.
Anatomic placement of the regions of interest was matched as closely as possible on T1-weighted and T2-weighted MR images. The signal intensity of the liver was recorded as the mean of three readings from regions of interest placed in the right and left lobes to account for signal heterogeneity. The signal intensity of the spleen was similarly measured with a region of interest that was 1–2 cm in diameter. The mean spleen signal intensity was calculated from three regions of interest obtained in the spleen in three sections. The standard deviation of the signal intensity measurements within each region of interest was kept to less than 10%. Liver fat was quantified on T1-weighted gradient-echo MR images as the percentage of relative signal intensity loss of the liver on out-of-phase images, with the following formula: (SIin – SIout)/SIin · 100, where SI is the mean liver signal intensity divided by the mean spleen signal intensity, SIin is in-phase signal intensity, and SIout is out-of-phase signal intensity.
Liver fat was quantified on T2-weighted fast spin-echo MR images as the percentage of relative signal intensity loss of the liver on fat-saturated images with the following formula: (SInonfat – SIfat)/SInonfat · 100, where SI was derived in the same manner as was denoted before, SInonfat is signal intensity without fat saturation, and SIfat is signal intensity with fat saturation.
The signal intensity of the spleen was used as a denominator in the formula to adjust for the lack of an objective signal intensity scale at MR imaging (8). The percentage of relative signal intensity losses on out-of-phase and fat-saturated MR images were considered reasonable measurements of liver fat on the basis of the known effect of fat on out-of-phase and fat-suppressed signal intensity values but were not considered to be a direct measure of histopathologically determined percentage of fat.
Histopathologic Analysis
Liver tissue was obtained for histopathologic analysis 2–12 weeks after MR imaging (mean and median, 8 and 9 weeks, respectively) in 25 patients and 2 weeks prior to MR imaging in two patients. This tissue was obtained with percutaneous core biopsy (n = 10), hepatectomy (n = 9), or partial resection (n = 8). One attending pathologist (S.K.) with 5 years of experience with liver histologic analysis reviewed the histopathologic slides for all patients without knowledge of imaging data. A second reader was not used because good interobserver agreement has been demonstrated for histopathologic liver fat determination (9). Researchers (10) in prior studies used a grading system for histopathologically determined liver fat as follows: grade 0, less than 5%; grade 1, 6%–33%; grade 2, 34%–66%; and grade 3, greater than 66%. The grading system incorporates the accepted normal value of liver fat, which is less than 5%. We used a more detailed technique with a continuous measurement of liver fat (0%, which indicated no hepatocytes with steatosis, to 100%, which indicated steatosis of all hepatocytes) rather than a large grouping of percentages of fat. Fat content was determined as the percentage of fat-containing hepatocytes in hematoxylin-eosin–stained specimens, as determined with standardized visual analysis (11).
Statistical Analysis
Statistical analysis was performed by using software (Stata, version 7.0; Stata, College Station, Tex). Linear regression analysis and the Pearson correlation coefficient (r) were used to examine the association between signal intensity loss on gradient-echo and spin-echo MR images and histopathologically determined percentage of fat. Comparisons were performed separately for patients without cirrhosis (n = 11) and patients with cirrhosis (n = 16). Correlation coefficients were compared by using the Hotelling t test (12). Differences with P values less than .05 were considered statistically significant.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Gradient-echo sequences lack the 180° refocusing pulses of spin-echo sequences, with consequent accumulation of loss of phase and rapid loss of transverse magnetization that result in signal intensity loss (15). The T2* effect of iron in cirrhosis could mask the presence of fat at out-of-phase MR imaging (16) and might explain why out-of-phase imaging was not found to be reliable for quantification of liver fat in patients with cirrhosis. T2-weighted fast spin-echo sequences are reported to have less susceptibility to the paramagnetic effects of iron than do gradient-echo sequences, and they have less sensitivity in the detection of mild increases in liver iron content. In addition, the images obtained without and with fat saturation were acquired with the same echo time so that any T2* effect should be equal for both sequences.
In concordance with findings in our study, those reported by Levenson et al (5) indicated a good correlation between signal intensity ratios on in-phase and out-of-phase gradient-echo MR images and histopathologically determined percentage of fat in noncirrhotic patients (n = 16, r = 0.86). Ultrasonography (US) and computed tomography (CT) have also been used to quantify liver fat. Saadeh et al (17) reported that moderate fatty infiltration (histopathologically determined percentage of fat of 33% or greater) could be reliably detected with sensitivity values of 100% and 93% but low positive predictive values of 62% and 76% for US and CT, respectively. Ricci et al (18) reported a correlation coefficient of 0.83 between the ratio of the density of the liver to that of the spleen at CT performed without contrast material enhancement and the percentage of liver fat, and this correlation coefficient is similar to that of in-phase and out-of-phase MR imaging in noncirrhotic patients. Limitations of US and CT include the inability to aid in distinguishing between fibrosis and fat with US and the masking effect of factors that increase liver density, such as the presence of iron, copper, glycogen, or amiodarone, with CT (16). Limanond et al (19) reported a correlation coefficient of 0.92 between the ratio of the density of the liver to that of the spleen with nonenhanced CT and the percentage of liver fat in potential liver donors (n = 42); however, in two of 27 patients, parenchymal hemosiderin resulted in an increase in liver attenuation so that it was within the normal range. In their study, four of the 27 patients with acceptable CT findings were considered poor liver donor candidates because core biopsy results revealed subtle hepatic necrosis and nonspecific hepatitis, and this discrepancy between these results highlights another limitation of CT evaluation of the liver.
In essence, T2-weighted fast spin-echo MR imaging without fat saturation and with fat saturation appears to provide the purest evaluation of fatty infiltration, without the confounding effects of T2*, fibrosis, and other factors that limit assessment with in-phase and out-of-phase MR imaging, US, or CT. The utility of T2-weighted sequences for liver fat quantification has been controversial. Kreft et al (7) reported a good correlation (r = 0.92) between absolute T2 values of the liver and the degree of liver fat infiltration on the basis of ex vivo experimental rat models. Levenson et al (5) reported that calculated T2 values showed no correlation with the degree of liver fat infiltration. Their calculations, however, were based on absolute T2 values within a region of interest in the liver rather than on the relative loss of signal intensity on fat-saturated T2-weighted MR images. In our study, percentage of relative signal intensity loss with T2-weighted sequences performed without fat saturation and with fat saturation was used for fat quantification, which may be more reliable than measurement of absolute T2 values of the liver.
Some limitations of our study should be noted. First, this was a retrospective study with small numbers of patients and without healthy control subjects. Second, liver biopsy and MR imaging were not contemporaneous, and alteration in the degree of liver fat may have occurred during the 3-month interval between performance of biopsy and MR imaging. Any such changes in liver fat, however, would have affected the results for both out-of-phase gradient-echo MR images and fat-saturated fast spin-echo MR images. Third, 10 of our histopathologic results were based on those obtained at percutaneous core biopsy; nine, on those obtained at hepatectomy; and eight, on those at partial liver resection. Although percutaneous core biopsy is the standard of reference in practice, results with this technique and with partial liver resection may not have been representative of those for the entire liver and may have introduced a sampling bias. A good correlation for histopathologically determined percentage of fat, however, has been reported between results obtained with paired percutaneous core biopsy performed on the right and left lobes of the liver (
= 0.81) (20). Fourth, since this was a retrospective study, we did not measure liver iron, although the presence of iron was speculated to be an explanation for the empirical observation of increased relative liver signal intensity on out-of-phase MR images in cirrhotic patients. Further studies in which a correlation is shown between relative signal intensity of the liver at MR imaging and histopathologically determined liver iron are required. Fifth, it is conceivable that an absolute excess of fat over water protons might result in decreased signal intensity loss on out-of-phase images. The histopathologically determined percentage of liver fat, however, reflects the percentage of fat-containing hepatocytes and not the percentage of fat composition of the liver tissue. It is not clear if the absolute number of fat protons can exceed the number of water protons in fatty infiltration of the liver. Sixth, the diagnosis of fatty liver is based on a percentage of greater than 5% of histopathologically determined liver fat (9), but the small numbers of patients in our study precluded establishment of criteria for the diagnosis of fatty liver by using MR imaging with this threshold value. Seventh, only seven patients were imaged by using dual in-phase and out-of-phase acquisitions (dual phase, n = 7). We used the spleen as an internal standard for all sequences, however, to account for differences in acquisition. Eighth, two different T2-weighted sequences were employed, fat-saturated fast spin-echo and single-shot fast spin-echo without fat saturation, with use of the spleen as an internal reference for both sequences. The sequences were part of the clinical protocol in this retrospective study, and despite the use of two different T2-weighted sequences, we found a significant correlation between the relative signal intensity loss on fat-suppressed T2-weighted images and the histopathologically determined percentage of fat. We believe our results are still valid, as both sequences were fast spin-echo sequences that employed multiple 180° refocusing pulses. Ninth, coronal T2-weighted MR images obtained without fat saturation were acquired in six patients. The regions of interest, however, were placed such that the segmental location, relationship to vessels, and depth within the liver were matched to those on the T2-weighted transverse images.
In conclusion, preliminary results suggest that liver fat may be more accurately quantified with fat-saturated fast spin-echo MR imaging than with out-of-phase gradient-echo MR imaging, especially in patients with cirrhosis.
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, J.S.G.; study concepts and design, J.S.G., A.Q., F.V.C.; literature research, J.S.G., A.Q., R.B.M.; clinical studies, J.S.G., A.Q., F.V.C.; data acquisition and analysis/interpretation, J.S.G., S.K.; statistical analysis, J.S.G., B.M.Y.; manuscript preparation, A.Q., J.S.G.; manuscript definition of intellectual content, editing, revision/review, and final version approval, all authors
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. P. O'Regan, M. F. Callaghan, M. Wylezinska-Arridge, J. Fitzpatrick, R. P. Naoumova, J. V. Hajnal, and S. A. Schmitz Liver Fat Content and T2*: Simultaneous Measurement by Using Breath-hold Multiecho MR Imaging at 3.0 T--Feasibility Radiology, May 1, 2008; 247(2): 550 - 557. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. G. Lall, A. M. Aisen, N. Bansal, and K. Sandrasegaran Nonalcoholic Fatty Liver Disease Am. J. Roentgenol., April 1, 2008; 190(4): 993 - 1002. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. C. A. Westphalen, A. Qayyum, B. M. Yeh, R. B. Merriman, J. A. Lee, A. Lamba, Y. Lu, and F. V. Coakley Liver Fat: Effect of Hepatic Iron Deposition on Evaluation with Opposed-Phase MR Imaging Radiology, February 1, 2007; 242(2): 450 - 455. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
