JOURNAL OF MAGNETIC RESONANCE IMAGING 28:246 –251 (2008)
Technical Note
Effects of Intravenous Gadolinium Administration and Flip Angle on the Assessment of Liver Fat Signal Fraction With Opposed-Phase and In-Phase Imaging Takeshi Yokoo, MD, PhD, Julie M. Collins, Robert F. Hanna, BS, Mark Bydder, PhD, Michael S. Middleton, MD, PhD, and Claude B. Sirlin, MD* Purpose: To assess the effects of intravenous gadolinium (Gd) and flip angle (FA) on liver fat quantification by opposed-phase (OP) and in-phase (IP) imaging.
Key Words: fatty liver disease; steatohepatitis; opposedphase imaging; in-phase imaging; fat quantification; gadolinium J. Magn. Reson. Imaging 2008;28:246 –251. © 2008 Wiley-Liss, Inc.
Materials and Methods: Our Institutional Review Board (IRB) approved this Health Insurance Portability and Accountability Act (HIPAA)-compliant, retrospective, clinical study. We identified 79 patients in whom abdominal OP and IP gradient-echoes were obtained at 1.5T before and after Gd administration. All 79 patients were imaged at high FA (ⱖ70°); 57 were also imaged at low FA (ⱕ20°). Fat signal fraction (FSF) was calculated from pre- and post-Gd liver images for each subject and FA using the formula, FSF ⫽ (SIP – SOP)/2SIP, where SIP and SOP are the OP and IP signal intensities, respectively. The dataset pairs (pre-Gd vs. post-Gd; high-FA vs. low-FA) were compared using linear regression analysis.
OPPOSED-PHASE (OP) AND IN-PHASE (IP) imaging are popular magnetic resonance (MR) techniques for examination of the abdomen. They are typically implemented as a T1-weighted gradient-echo sequence in which images are acquired at two echo times (TEs), at which the fat and water signals are assumed to be OP and IP, respectively. This technique is useful for evaluation of fatty liver disease (1– 6); tissue fat content can be quantified by fat signal fraction (FSF), calculated from the pair of OP and IP signal intensities as,
Results: Before Gd, FSF was significantly greater at high FA than at low FA, with regression parameters (slope/intercept) of 1.27*/0.02*, where * indicates P value ⬍0.01. After Gd, FSF was similar at high and low FA (0.99/– 0.00). Gd administration caused an FA-dependent reduction in FSF, larger at high FA (0.68*/– 0.03*) than at low FA (0.94, – 0.01*).
FSF ⫽ 共S IP ⫺ S OP 兲/2S IP ,
Conclusion: FSF by OP-IP imaging is highly dependent on FA before Gd, but this dependency is eliminated after administration of Gd. Gd appears to minimize the effect of T1-weighting and may improve the accuracy of liver fat quantification.
Department of Radiology, University of California, San Diego, San Diego, California, USA. Contract grant sponsor: National Institutes of Health (NIH); Contract grant numbers: U01 DK61734; R01 DK075128-01 *Address reprint requests to: C.B.S., MD, Department of Radiology, University of California, San Diego, 408 Dickinson St., San Diego, CA 92122-8226. E-mail:
[email protected] Received September 10, 2007; Accepted February 12, 2008. DOI 10.1002/jmri.21375 Published online in Wiley InterScience (www.interscience.wiley.com).
© 2008 Wiley-Liss, Inc.
[1]
where SIP is the IP signal intensity, and SOP the OP signal intensity. Tissue T1 relaxation time and T1-weighting parameters, e.g., flip angle (FA) and repetition time (TR), can influence the observed FSF values (5,7–9). Although not previously reported in liver, low molecular weight gadolinium (Gd) chelates could also influence FSF due to their relaxation altering effects. In lipid-containing phantoms and in vivo lipid-rich tissues, Gd has been shown to increase the water signal relative to the fat signal (10), thereby lowering the calculated FSF. Based on this observation, we hypothesized that Gd administration would reduce the calculated FSF of the liver in human subjects. We further hypothesized that the Gdmediated reduction would vary with the T1-weighting of the sequence and thus could depend on the FA. The purpose of this study was to assess the effects of Gd administration and FA on liver FSF by OP-IP imaging. Since routine liver MR exams today often involve Gd administration, it would be important to understand the relationship between Gd and OP-IP imaging,
246
Effects of Gd and FA on Liver Fat Quantification
247
Table 1 Clinical Indications for Intravenous Gadolinium Administration in Study Population Clinical indications for Gd Extrahepatic malignancy Diffuse liver disease Liver nodule Extrahepatic mass Other Total
Number of patients 29 25 17 5 3 79
(10°–20°) FAs. Other parameters included slice thickness ⫽ 6 – 8 mm, receiver bandwidth ⫽ 128 kHz, and matrix size ⫽ 256 ⫻ 160 to 256 ⫻ 256. The field of view was adjusted for each patient’s body habitus. All 79 subjects were imaged at high FA; 57 of the subjects were also imaged at low FA. Intensity images were stored on picture archiving and communication system (PACS) for further analysis. Sequences other than the OP-IP acquisition were not analyzed in this study. Image Analysis
especially in practices for which a high prevalence of fatty liver disease is expected. MATERIALS AND METHODS Patients Our Investigational Review Board (IRB) approved this Health Insurance Portability and Accountability Act (HIPAA)-compliant, retrospective, cross-sectional study. Between January 2006 and April 2007, the standard clinical protocol for abdominal MRI at our institution included OP-IP gradient-echo imaging at a high FA (ⱖ70°). For patients with known or suspected fatty liver and patients with focal nodules, a low FA (ⱕ20°) sequence was added at the discretion of the monitoring radiologist to better assess fat content (5,11); if Gd was given for clinical care, the same OP-IP sequence(s) was repeated after Gd administration. During this study period, 174 patients underwent MR examinations at one of our outpatient MR centers, received intravenous Gd, and were imaged with OP-IP gradient-echo sequences. We excluded 93 patients in whom imaging parameters were not identical on pre-Gd and post-Gd OP-IP images and one patient with biopsyconfirmed hemosiderosis, as the calculated FSF is known to be unreliable in iron-overloaded liver (5,6,12). The remaining 79 patients formed the study group. Patients had a mean age of 55.9 years (range ⫽ 18 – 89; 35 males, 44 females). A total of 32 patients had a history of chronic liver disease (viral hepatitis, 17; alcoholic fatty liver disease, three; nonalcoholic fatty liver disease, six; and other causes, six). The clinical indications for the MR examinations are summarized in Table 1. Imaging Imaging examinations were performed on a 1.5T Siemens Symphony whole-body clinical system (Siemens Medical Systems, Erlangen, Germany). Patients were positioned supine with a phased-array coil centered over the abdomen. Single-breathhold axial 2D interleaved dual-echo spoiled gradient recalled echo (GRE) images were obtained before and 4 –10 minutes after intravenous administration of gadobenate dimeglumine (MultiHance; Bracco Diagnostics Inc, Princeton, NJ, USA), at a dose ⫽ 0.2 mL/kg, with TR ⫽ 101–122 msec, TEs ⫽ 2.30 –2.38 msec (OP) and 4.60 – 4.76 msec (IP), at high FA (70°–90°) only or at both high and low
Images were evaluated qualitatively (in consensus by C.B.S., a senior body-imaging attending physician, and T.Y., a radiology resident) and quantitatively. For the quantitative analysis, a trained research technologist (J.M.C., with three years of experience) selected a representative transverse slice of the liver on each subject and manually colocalized the slice on all available dualecho images. Avoiding organ boundaries, imaging artifacts, major vessels, and bile ducts, a single elliptical region of interest (ROI) ranging in size from 124 to 160 mm2 was drawn within the liver on the pre-Gd OP image at high FA. Using PACS software, the ROI was automatically propagated to the other colocalized images. The registration of the ROI was visually checked and adjusted as needed by the research technologist. For each pair of IP and OP images, the average signal intensity within the registered ROI was recorded, and the FSF values were calculated using Eq. [1]. In select patients, FSF maps were generated for illustrative purposes, computed from the source OP and IP images by applying Eq. [1] pixel by pixel. Statistical Analysis Using MATLAB™ (The Mathworks, Natick, MA, USA), two statistical analyses were performed on the FSF values: 1) comparison of pre-Gd vs. post-Gd; and 2) comparison of low-FA vs. high-FA. In the first analysis, pre-Gd (x-axis) vs. post-Gd (yaxis) FSF values were plotted on a scatter diagram at each FA (high or low). In the second analysis, low-FA (x-axis) vs. high-FA (y-axis) FSF values were plotted on a scatter diagram for each contrast status (pre- and post-Gd). Using linear regression models, the leastsquare linear fits were drawn through the plotted values, and the slope and intercept estimates, as well as their standard errors, were calculated. The null hypothesis, that Gd administration (first analysis) or FA (second analysis) did not affect the calculated FSF, was represented by a diagonal line with unit slope and zero intercept. Any systematic deviation of the data’s distribution from the diagonal line (as manifested by a regression slope not equal to 1 or a regression intercept not equal to 0) would suggest that Gd (first analysis) or FA (second analysis) affected FSF. To assess statistical significance, the best-fit slope and intercept were subjected to Student’s t-test against the null hypothesis of unit slope and zero intercept, respectively. A calculated P value less than 0.01 was considered significant.
248
Yokoo et al.
high-FA data had best-fit slope of 0.68 (P value ⬍ 0.001) and intercept of – 0.024 (P ⬍ 0.001). The low-FA data had slope of 0.94 (P ⫽ 0.017) and intercept of – 0.006 (P ⫽ 0.009). Figure 4 shows the results of the regression analyses of FSF values comparing high-FA vs. low-FA data (N ⫽ 57) before (Fig. 4a) and after (Fig. 4b) Gd administration. Before Gd, the data had best-fit slope of 1.27 (P ⬍ 0.001) and intercept of 0.020 (P ⬍ 0.001). By comparison, after Gd, the data had best-fit slope of 0.99 (P ⫽ 0.33) and intercept of – 0.002 (P ⫽ 0.11). DISCUSSION
Figure 1. OP and IP gradient-echo images before (left panels) and after (right panels) Gd administration of a 62-year-old woman with fatty liver. Signal intensities in the high FA images (top row) are approximately twice those in the low FA images (bottom row). Compared to the IP images, the signal intensity of the liver is lower on the OP images. The degree of OP signal loss is greatest for the pre-Gd pair of images acquired at high FA. FA ⫽ flip angle, Gd ⫽ gadolinium.
RESULTS Qualitative Analysis of Images In general, high-FA images had greater apparent signalto-noise ratio and generated higher soft-tissue contrast than low-FA images, regardless of whether images were obtained pre- or post-Gd administration (Fig. 1; top vs. bottom). At high FA, Gd administration noticeably increased the signal intensity of liver on both OP and IP images; at low FA, the effects of Gd on liver signal intensity were not obvious, regardless of whether images were OP or IP (Fig. 1; left vs. right). In patients with fatty liver, there was noticeable signal loss on the OP image compared to the IP image. Prior to Gd administration, the OP-IP intensity difference was visibly greater at high than at low FA. After Gd administration, the OP-IP intensity difference was smaller, and it was comparable to that observed at low FA without Gd administration. Thus, Gd administration reduced the OP-IP signal intensity difference and the discrepancy between low-FA and high-FA images. In keeping with these observations, Gd administration reduced the intensity of the corresponding FSF maps and the difference between low-FA and high-FA maps (Fig. 2). Although post-Gd FSF maps at high FA were similar to the maps at low FA in the pixel values, they were noticeably less noisy.
OP and IP gradient-echo imaging is routinely performed in clinical liver MR examinations. While useful for assessment of fatty liver, the apparent FSF has been shown to vary with T1-weighting of the imaging sequence, such as FA (5,11,13,14). As predicted by gradient-echo signal equation, high T1-weighting preferentially suppresses the long T1 signal of water (T1 ⬇ 570 msec (15)) relative to the short T1 signal of fat (T1 ⬇ 200 msec (16)), and causes FSF overestimation in proportion to the underlying tissue fat content. To minimize this “T1-bias,” Liu et al (11) proposed the use of a low-FA sequence (low T1-weighting) and showed that this improved fat quantification accuracy in an oil-water mixture phantom. Our study suggests that Gd administration may also help minimize the T1-bias. As expected, we found that FSF values were proportionally overestimated (regression slope ⬎1) at high FA compared to low FA (Fig. 4a) before Gd administration. After Gd administration there was no overestimation; FSF values were statistically indistinguishable between high and low FAs (Fig. 4b). This debiasing effect of Gd was also seen in Fig. 3a as significant reduction in the proportionality (slope ⬍1) at high FA, corresponding to the reversal of the FSF
Quantitative Analysis of Images Depending on the acquisition technique (pre- or postGd; high- or low-FA), the range of calculated FSF in the study population sample was –12% to 36% and negative values were seen in 35– 44 out of 79 patients, likely due to T2* effects. Figure 3 shows the results of the regression analyses comparing pre-Gd vs. post-Gd FSF values at high FA (Fig. 3a; N ⫽ 79) and at low FA (Fig. 3b; N ⫽ 57). The
Figure 2. FSF maps of the patient in Fig. 1, computed at high (top) and low (bottom) FAs by applying Eq. [1] pixel-by-pixel to the pre-Gd (left) and post-Gd (right) images. The average FSF value (expressed as a percentage) within an ROI (not shown) is shown for each acquisition method. The pre-Gd high-FA acquisition method predicts the highest FSF. The other three methods predict a lower FSF and are in close agreement with each other. FA ⫽ flip angle, Gd ⫽ gadolinium.
Effects of Gd and FA on Liver Fat Quantification
249
tributed to our empirical observations. The effect of T2* relaxation on OP-IP imaging is well documented (5,6,12). In this study at 1.5T, two consecutive OP and IP echoes were acquired 2.3 msec apart. During this interecho interval, fat and water signals would undergo T2* decay; this signal loss would confound any signal change due to phase interference between OP and IP echoes, causing systematic FSF underestimation. In particular, excessive iron deposition such as in hemosiderosis can cause significant T2* shortening below 10 msec (normal liver 25–30 msec in our scanner) and FSF can be grossly underestimated. While we excluded from this study patients with known or suspected iron overload, some underestimation would be expected due to the inherent T2* of liver. Since T2*-bias affects FSF
Figure 3. Pre-Gd (x-axis) vs. post-Gd (y-axis) FSFs calculated using high FA (a: N ⫽ 79) and low FA (b: N ⫽ 57) datasets. The solid line is the best-fit line through the data (linear regression). The dotted line has slope ⫽ 1 and intercept ⫽ 0 and represents the null hypothesis. At high FA, Gd administration significantly reduces the apparent FSF. The reduction has scalar (slope) and fixed (intercept) components. At low FA, the effect of Gd is less pronounced. *P ⬍ 0.01; FA ⫽ flip angle, Gd ⫽ gadolinium.
overestimation. In Fig. 3b, Gd caused no significant change in the proportionality (slope ⬇ 1), most likely because the T1-bias had already been minimized by the use of low FA. This observed Gd effect and its modulation by T1-weighting (i.e., FA) are consistent with the known property of Gd as primarily a T1-shortening agent. Based on these data, we hypothesize that Gd may have caused: 1) preferential T1 shortening of water over fat, such that the post-Gd T1s were relatively similar; or 2) shortening of both T1s such that both signals were more recovered by the next TR, or a combination of both. Our study was not designed to elucidate the mechanism of Gd action, and further research would be necessary to confirm this hypothesis. While the main Gd effect appears to be related to T1 shortening, Gd could also have shortened T2* and con-
Figure 4. Low FA (x-axis) vs. high FA (y-axis) FSFs calculated using pre-Gd (a: N ⫽ 57) and Post-Gd (b: N ⫽ 57) datasets. The solid line is the best-fit line through the data (linear regression). The dotted line has slope ⫽ 1 and intercept ⫽ 0 and represents the null hypothesis. Pre-Gd, FSF values are significantly higher at high FA than low FA. The discrepancy has scalar (slope) and fixed (intercept) components. Post-Gd, high-FA and low-FA values are similar. *P ⬍ 0.01; FA ⫽ flip angle, Gd ⫽ gadolinium.
250
irrespective of the amount of underlying fat, lean livers with little or no fat signals can also be underestimated, resulting in the negative FSF values as seen in Figs. 3 and 4. The small but statistically significant negative regression intercepts in Fig. 3 may represent Gd-mediated T2* shortening, as worsening of the T2*-bias in the post-Gd dataset is expected to cause downward displacement of the regression lines. While the extent of T2* shortening by Gd was not quantified in this study, its effect was presumably small, because grossly shortened post-Gd T2* would have caused profound negative displacement of the regression lines at both FAs. Specialized T2* estimation and correction techniques using multiple-echo acquisition protocols (5,9) have been proposed recently, which may reduce or eliminate T2*bias. In this study using standard OP-IP imaging, however, T2* correction was not possible due to insufficient data points (⬎2 echoes necessary) for simultaneous fat and T2* quantification. Recent phantom studies suggested that correction of both T1- and T2*-biases may be necessary for accurate fat quantification by FSF (9,11). As previously mentioned, the T1-bias could be corrected using a low FA to minimize the T1-weighting, and the T2*-bias using a multiecho acquisition sequence to estimate tissue T2*. This study suggests Gd-enhanced imaging may be an alternative method for T1-bias correction. However, we advise caution against routine administration Gd for the sole purpose of fat quantification. Gd has rare but significant risks, including anaphylaxis and nephrogenic systemic fibrosis (17). Although presumed small, Gd may also exacerbate FSF underestimation by T2* shortening, and may necessitate additional T2* estimation and correction methods to preserve total accuracy. This study had the following limitations not previously mentioned. There was no definitive reference standard for the tissue fat content and the total FSF accuracy could not be directly assessed. While the pre-Gd low-FA data may serve as the T1-unbiased reference standard, further investigation is still required to validate the results of the oil-water phantom study (11) in vivo. We assessed the post-Gd FSF in the delayed phase (4 –10 minutes after a bolus injection) when the Gd concentration was presumed to be at equilibrium with the interstitium. The timing of the imaging was not strictly controlled, and so the calculated FSF could have varied dynamically with temporal changes in Gd distribution. However, based on the tight linear clustering seen in the pre- vs. post-Gd analyses (Fig. 3), the timing of imaging was probably not a critical factor. We used gadobenate dimeglumine as a prototypical Gd agent for liver imaging, but different Gd formulations could have slightly different effects. Our preliminary data using other Gd agents, such as gadoversetamide and gadopentetate, suggest that the qualitative effect is quite similar across different formulations. We maximized the sample size by grouping patients who were imaged with slightly different T1-weighting (TR ⫽ 101– 122 msec; FA ⫽ 10 –20° vs. 70 –90°). The tight distribution of the data within each FA group (Fig. 3) and distinct separation between two FA groups (Fig. 4a) suggest that the slight variability in the T1-weighting was inconsequential. The limitations of the OP-IP tech-
Yokoo et al.
nique itself, such as the fat-water dominance ambiguity (5,18) and complexity of fat-water interference (19,20) are outside of the scope of this work but are discussed in detail elsewhere. While these represent important conceptual limitation of this technique, their clinical significance is still unclear, and they are not expected to alter the observed Gd effect in this study. In conclusion, this retrospective clinical study of 79 patients showed that Gd administration in OP-IP imaging of the liver eliminated the T1-weighting dependency of FSF by suppressing the overestimation caused by high FA. While this study suggests a potential new role of Gd for assessment of liver fat, we advise caution against its routine use for the sole purpose of fat quantification due to the risks associated with Gd administration.
REFERENCES 1. Dixon WT. Simple proton spectroscopic imaging. Radiology 1984; 153:189 –194. 2. Schertz LD, Lee JK, Heiken JP, Molina PL, Totty WG. Proton spectroscopic imaging (Dixon method) of the liver: clinical utility. Radiology 1989;173:401– 405. 3. Heiken JP, Lee JK, Dixon WT. Fatty infiltration of the liver: evaluation by proton spectroscopic imaging. Radiology 1985;157:707– 710. 4. Levenson H, Greensite F, Hoefs J, et al. Fatty infiltration of the liver: quantification with phase-contrast MR imaging at 1.5 T vs biopsy. AJR Am J Roentgenol 1991;156:307–312. 5. Hussain HK, Chenevert TL, Londy FJ, et al. Hepatic fat fraction: MR imaging for quantitative measurement and display— early experience. Radiology 2005;237:1048 –1055. 6. Qayyum A, Goh JS, Kakar S, Yeh BM, Merriman RB, Coakley FV. Accuracy of liver fat quantification at MR imaging: comparison of out-of-phase gradient-echo and fat-saturated fast spin-echo techniques—initial experience. Radiology 2005;237:507–511. 7. Hughes F, Bydder M, Middleton MS, et al. Effects of T1 and T2 relaxation on liver fat quantification using GRE sequences. AJR Am J Roentgenol 2006;186:A57–A60. 8. Fishbein MH, Gardner KG, Potter CJ, Schmalbrock P, Smith MA. Introduction of fast MR imaging in the assessment of hepatic steatosis. Magn Reson Imaging 1997;15:287–293. 9. Yu H, McKenzie CA, Shimakawa A, et al. Multiecho reconstruction for simultaneous water-fat decomposition and T2* estimation. J Magn Reson Imaging 2007;26:1153–1161. 10. Mitchell DG, Stolpen AH, Siegelman ES, Bolinger L, Outwater EK. Fatty tissue on opposed-phase MR images: paradoxical suppression of signal intensity by paramagnetic contrast agents. Radiology 1996;198:351–357. 11. Liu CY, McKenzie CA, Yu H, Brittain JH, Reeder SB. Fat quantification with IDEAL gradient echo imaging: correction of bias from T(1) and noise. Magn Reson Med 2007;58:354 –364. 12. Westphalen AC, Qayyum A, Yeh BM, et al. Liver fat: effect of hepatic iron deposition on evaluation with opposed-phase MR imaging. Radiology 2007;242:450 – 455. 13. Bydder M, Middleton MS, Chaves AD, Sirlin CB. Effect of flip angle on fat quantification by Dixon techniques. In: Proceedings of the 14th Annual Meeting of ISMRM, Seattle, WA, USA, 2006 (Abstract 2300). 14. Fishbein MH, Stevens WR. Rapid MRI using a modified Dixon technique: a non-invasive and effective method for detection and monitoring of fatty metamorphosis of the liver. Pediatr Radiol 2001; 31:806 – 809. 15. Bottomley PA, Foster TH, Argersinger RE, Pfeifer LM. A review of normal tissue hydrogen NMR relaxation times and relaxation mechanisms from 1-100 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med Phys 1984;11:425– 448. 16. Bluml S, Schad LR, Stepanow B, Lorenz WJ. Spin-lattice relaxation time measurement by means of a TurboFLASH technique. Magn Reson Med 1993;30:289 –295.
Effects of Gd and FA on Liver Fat Quantification 17. Collidge TA, Thomson PC, Mark PB, et al. Gadolinium-enhanced MR imaging and nephrogenic systemic fibrosis: retrospective study of a renal replacement therapy cohort. Radiology 2007;245:168 – 175. 18. Reeder SB, Pineda AR, Wen Z, et al. Iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL): application with fast spin-echo imaging. Magn Reson Med 2005;54:636 – 644.
251 19. Brix G, Heiland S, Bellemann ME, Koch T, Lorenz WJ. MR imaging of fat-containing tissues: valuation of two quantitative imaging techniques in comparison with localized proton spectroscopy. Magn Reson Imaging 1993;11:977–991. 20. Bydder M, Yokoo T, Hamilton G, Middleton MS, Chavez AD, Sirlin CB. Fat-fat interactions in Dixon-variant imaging. In: Proceedings of the 15th Annual Meeting of ISMRM, Berlin, Germany, 2007 (Abstract 1632). [Joint Annual Meeting of ISMRM-ESMRMB].
