PRECLINICAL AND CLINICAL IMAGING Notes
Magnetic Resonance in Medicine 73:1602–1608 (2015)
Assessment of Diffusion Tensor MR Imaging (DTI) in Liver Fibrosis with Minimal Confounding Effect of Hepatic Steatosis Yunjung Lee1 and Hyeonjin Kim1,2* Purpose: Given the potential confounding effect of fat on apparent diffusion coefficient (ADC) in the liver, we have assessed diffusion tensor imaging in liver fibrosis with minimal effect of fat on ADC and fractional anisotropy (FA). Methods: Thirty-six mice were used, among which 20 mice were CCl4 treated for fibrosis induction. Diffusion tensor imaging was performed at 9.4T using a spin-echo diffusion tensor imaging sequence with six gradient directions. Hepatic fat fraction obtained by MR spectroscopy was used as hepatic fat content. Fibrosis scores were obtained from histopathology. Results: The hepatic fat fractions of the two animal groups were below 5.5% and not different (5.3 6 1.5 vs. 4.6 6 1.1%; P ¼ 0.115). Fibrosis scores were higher in CCl4-treated mice (0.0 6 0.0 vs. 2.1 6 0.7; P < 0.001). Nonetheless, there was no difference in ADC between the two groups (0.711 6 0.068 103 vs. 0.718 6 0.095 103 mm2 s1; P ¼ 0.911). The treated group had a lower FA than control (0.552 6 0.050 vs. 0.586 6 0.013; P ¼ 0.023). ADC was not correlated with hepatic fat fraction and fibrosis. FA was correlated with hepatic fat fraction (r ¼ 0.418, P ¼ 0.011) and fibrosis (r ¼ 0.411, P ¼ 0.012). Conclusion: FA may be more sensitive to mild-to-moderate liver fibrosis than ADC. In addition to ADC, FA may also be sensitive to hepatic fat content, and therefore need careful interpretation in liver fibrosis with concomitant fatty liver. Magn Reson C 2014 Wiley Periodicals, Inc. Med 73:1602–1608, 2015. V Key words: diffusion magnetic resonance imaging; diffusion tensor imaging; liver fibrosis; apparent diffusion coefficient; fractional anisotropy; steatosis
INTRODUCTION The performance of diffusion weighted magnetic resonance imaging has been extensively explored in diseased livers (1–9), and apparent diffusion coefficient (ADC) has been suggested to be a potential MR parame1
Department of Radiology, Seoul National University Hospital, Seoul, Korea. Department of Biomedical Sciences, Seoul National University, Seoul, Korea. Grant sponsor: The Ministry of Education, Science and Technology (Basic Science Research Program through the National Research Foundation of Korea [NRF]); Grant numbers: 2009–0077642, 2010–0002896, 2013R1A1A2013516. *Correspondence to: Hyeonjin Kim, Ph.D., Department of Radiology, Seoul National University Hospital, 101 Daehangno, Jongnogu, Seoul 110–744, Korea. E-mail: [email protected]
Received 2 January 2014; revised 10 March 2014; accepted 25 March 2014 DOI 10.1002/mrm.25253 Published online 14 April 2014 in Wiley Online Library (wileyonlinelibrary. com). 2
C 2014 Wiley Periodicals, Inc. V
ter in the diagnosis of liver diseases such as liver fibrosis (1,3,5) and inflammation (3,7). Recently, however, attention has been made to the potential confounding effects of fat on the estimation of ADC in the liver (6,10–15). For instance, several studies demonstrated that ADC can be altered not only by fibrosis and inflammation but by steatosis as well due to the restricted water diffusion arising from fat infiltration (6,10,13,14) and/or due to the presence of residual fat signal (6,12,15). On the other hand, no significant impact of steatosis on hepatic ADC measurement has also been documented (16,17). Although not fully in agreement, these previous studies emphasize that ADC values in the liver need to be carefully interpreted in the presence of hepatic steatosis. In addition to those studies above that focused on the relationship between ADC and fat, it should also be important to clarify the relationship between ADC and the severity of liver diseases such as liver fibrosis after accounting for the potential confounding effect of steatosis. However, it is challenging to address such issues due to several difficulties. First of all, fatty liver is common in liver patients, and patient cohorts can be heterogeneous. Moreover, the limited scan time in human abdominal study necessitates the use of an echo planar imaging (EPI) sequence for diffusion MRI, which is highly subject to image artifacts due to field inhomogeneity that is far more problematic for a large field of view of the abdomen (12). The extent of image degradation with EPI can be exacerbated in fatty liver due to severe chemical shift artifact along the phase-encoding direction resulting from imperfect fat suppression. In an effort of minimizing these difficulties, an ex vivo animal study has been reported recently where the relationship between ADC and disease progression was investigated on excised fibrotic liver specimens with minimal steatosis (18). Meanwhile, the performance of diffusion tensor imaging (DTI) in the assessment of liver diseases has also been studied by several groups (3,7,11,19), and, in addition to ADC, the potential role of fractional anisotropy (FA) in assessing diseased livers has been demonstrated (11,19). To this end, we have investigated the efficacy of DTI in the assessment of liver fibrosis in vivo in an experimental setting where the potential confounding effect of fat on ADC and FA can be negligible. For this purpose, a suitable animal model was used, in which a broad range of the severity of liver fibrosis was obtained and yet with minimal fat content throughout the course of the evolution of the disease. In addition, diffusion MR images
DTI in Liver Fibrosis with Minimal Steatosis
were collected in combination with a conventional spinecho (SE) sequence instead of EPI to minimize potential errors resulting from image artifacts. METHODS Animal Preparation The animal research protocol was approved by the Institutional Animal Care and Use Committee. Based on the previous reports (20), male C57BL/6 mice were used in this study (n ¼ 36). Mice were 6 weeks of age at the onset of the study. Liver fibrosis was induced in 20 mice by an intraperitoneal injection of carbon tetrachloride (CCl4) mixed with vegetable oil (25 mL CCl4 in a 150 mL volume [1:6]) at the frequency of three times per week (20–24), for 2–10 weeks [2 weeks (n ¼ 3), 3 weeks (1), 4 weeks (2), 5 weeks (2), 6 weeks (2), 7 weeks (2), 8 weeks (2), 9 weeks (3), and 10 weeks (3)]. There were 16 control mice, among which eight mice received pure vegetable oil intraperitoneally at the same frequency. Diffusion MRI and Single-Voxel Proton Magnetic Resonance Spectroscopy Mice underwent MR scans 2 days after the last dose of CCl4 or oil alone in control animals. Mice were anesthetized with isofluorane (1% in 100% oxygen). All MR data were collected on a 9.4T animal MR scanner (Bruker Biospec 94/20 USR; Bruker Biospin, Ettlingen, Germany). A transmit-only volume coil was used for RF transmission and a phased-array four channel surface coil was used for signal reception (Bruker Biospin, Ettlingen, Germany, for both coils). Scout images were acquired in axial, coronal, and sagittal planes using a gradient echo sequence. The imaging parameters for coronal and sagittal images were FOV ¼ 33 22 mm2, matrix size ¼ 256 256, repetition time (TR)/echo time (TE) ¼ 295/3 ms, flip angle ¼ 30 , number of slices ¼ 10 for coronal and 20 for sagittal images, slice thickness (TH) ¼ 1 mm, and one signal average. For axial images, they were FOV ¼ 33 22 mm2, matrix size ¼ 256 256, TR/TE ¼ 250/4.6 ms, flip angle ¼ 40 , number of slices ¼ 8, TH ¼ 1 mm, and four signal averages. DTI images were acquired in the axial plane using a respiratory-gated, fat-saturated, SE DTI sequence with six gradient directions (25) at b ¼ 500 s mm2. Images at b ¼ 0 s mm2 were also acquired. The duration and separation of diffusion gradients were 5 and 8.5 ms, respectively. Other imaging parameters were FOV ¼ 28 22 mm2, matrix size ¼ 256 128 (pixel resolution ¼ 0.109 0.172 mm2), TR/TE ¼ 4000/20 ms, number of slices ¼ 8, TH ¼ 1 mm, one signal average, and bandwidth ¼ 100 kHz. To minimize potential sampling errors with liver biopsy in the estimation of steatosis, proton magnetic resonance spectroscopy (1H-MRS) was performed for fat quantification, and the resulting hepatic fat fraction (HFF) was used as a measure of hepatic fat content instead of steatosis scores from histopathology. MRS spectra were collected using a stimulated echo acquisition mode sequence (26) with an outer-volume suppression module, which is preferentially used at high field
for short TEs attainable and large RF excitation bandwidth minimizing the voxel-shift effect (27). A hermite pulse of 0.6 ms duration and 9000 Hz bandwidth was used for all three 90 pulses (TR/TE/mixing time ¼ 5000/2.2/3 ms, spectral width ¼ 5000 Hz, number of data points ¼ 2048, 4 dummy scans, 32 signal averages). No water suppression was performed. To account for potential liver heterogeneity, spectra were acquired from a total of three voxels (1.3 1.3 1.3 mm3) for each animal. Based on the scout images, these voxels were carefully placed within the liver avoiding major blood vessels. All data were collected by a single operator to maintain the locations of the MRS voxels and MRI slices in the liver across all samples. Histopathology Following the MR scans, livers were harvested for histopathologic examination. Two to four 4-mm-thick pieces of liver from each mouse were fixed in 10% neutral buffered formalin, processed, embedded in paraffin, sectioned at 5 mm and stained with hematoxylin and eosin, and Masson’s trichrome by routine methods. Livers were scored for necrosis and inflammation by examining hematoxylin and eosin-stained slides, and for fibrosis by examining Masson’s trichrome-stained slides. Severity scores ranged from 0 to 4 (23,24). MR Data Analysis DTI data were analyzed using Matlab (v7.13; Mathworks Inc., Natick, MA) as previously described (28,29). Briefly, for each pixel, a diffusion vector composed of the six independent components (Dxx, Dyy, Dzz, Dxy, Dxz, and Dyz) of the diffusion tensor was calculated by multiplying the inverse of a 6 6 diffusion gradient matrix that is dependent exclusively on the diffusion gradient directions, and an ADC vector composed of the ADC values along the six different gradient directions. The diffusion tensor obtained thereafter was diagonalized to extract the three eigenvalues. Finally, mean diffusivity (one-third of the trace of the diffusion tensor) and FA were calculated according to the definitions given in Refs. (25) and (29), and the mean diffusivity was used as ADC (30). For each animal, regions of interest (ROIs) were defined on the image obtained at b ¼ 0 s mm2 for each slice (typically 3 ROIs per slice) including as much liver tissue as possible while avoiding major blood vessels (Fig. 1), and then the mean ADC and FA across the slices were used in the data analysis. The measurement reproducibility with the use of such arbitrary-shaped ROIs was tested by defining ROIs twice for all slices of the animals and estimating percent changes for ADC and FA between the two measurements. The water-unsuppressed MRS data were processed using MRUI (v3.0; (31)). First, data were Fouriertransformed, line-broadened (5 Hz), and phasecorrected. Then, the individual peak areas of water (4.7 ppm) and lipid (methylene at 1.3 ppm) resonances were estimated using AMARES (32). Finally, HFF was calculated by fat/(water þ fat) 100 for each voxel, and the mean HFF over the three voxels was used in the data
Lee and Kim
FIG. 1. Representative diffusion weighted images and diffusion MRI parameters along with the liver sections for a control (a–f) and a treated (g–l) mice. a and g: images at b ¼ 0 s mm2, b and h: images at b ¼ 500 s mm2 with the gradient direction of (x ¼ 1, y ¼ 1, z ¼ 0), c and i: ADC maps of the liver tissue, d and j: FA maps of the liver tissue, e and k: hematoxylin and eosin-stained liver sections, f and l: Masson’s trichrome-stained liver sections. Typical measurement ROIs are also shown in (a) and (g). Note the moderate amount of liver fibrosis (blue/gray) in (l) for the treated mouse in contrast to the control mouse (f). Nonetheless, fatty change is absent in (k) for the treated mouse as in (e) for the control mouse (control mouse: steatosis score ¼ fibrosis score ¼ 0; treated mouse: steatosis score ¼ 0, fibrosis score ¼ 2).
analysis. As the TE used herein was as short as 2.2 ms, we have not considered T2 corrections. Statistical Analysis All statistical analyses were performed using Matlab. All results were expressed as mean 6 standard deviation (SD). For pair-wise group comparisons, the Mann–Whitney U-test was used. For multiple pair-wise group comparisons Bonferroni correction was performed by adjusting the significance level (a). Otherwise, a P-value of less than 0.05 was considered to indicate statistical significance. The Spearman Rho test was used to examine correlations between the histopathologic and MRI parameters. RESULTS Figure 1 shows representative liver images at b ¼ 0 and 500 s mm2, and ADC- and FA-maps along with the hematoxylin and eosin- and Masson’s trichrome-stained liver sections for a control and a treated mice. The high signal-to-noise ratio, lack of artifacts, and high resolution of the images resulting from the use of SE instead of EPI are clearly demonstrated. From the measurement reproducibility test with repeated ROI definitions, the percent changes for all slices of the animals ranged 1.22–2.53% for ADC and 1.45– 1.59% for FA. The correlation coefficients between the two measurements for ADC and FA were 0.998 and 0.993, respectively (P < 0.001 for both). Control Mice with and without Oil Treatment There was no difference between the two control groups with and without oil treatment in age (13.4 6 2.3 vs. 11.5 6 2.4 weeks; P ¼ 0.162), weight (22.1 6 3.4 vs. 22.5 6 1.4 g; P ¼ 0.518), HFF (5.5 6 1.9 vs. 5.1 6 1.1 %; P ¼ 0.903), necrosis score (0.3 6 0.5 vs. 0.4 6 0.7; P ¼ 1.000),
inflammation score (0.0 6 0.0 vs. 0.1 6 0.4; P ¼ 1.000), and fibrosis score (0 for all control animals). None of the differences in the DTI parameters between the two control groups reached statistical significance (ADC ¼ 0.730 6 0.130 103 vs. 0.693 6 0.145 103 mm2 s1, P ¼ 0.505; FA ¼ 0.591 6 0.018 vs. 0.581 6 0.003, P ¼ 0.442). Therefore, the two control groups with and without oil treatment were grouped together as a control group (n ¼ 16). Treated Mice Grouped According to Fibrosis Scores Among the CCl4-treated 20 mice, there were 4, 11, and 5 mice with a fibrosis score of 1–3, respectively. The comparison among the control group and those three animal groups (F1–F3) stratified according to the fibrosis scores is shown in Figure 2. The mean HFF values were below 5.5% for all animal groups. On the other hand, as the fibrosis scores increased, the mean necrosis (Fig. 2b) and inflammation (Fig. 2c) scores also substantially increased. Note in Figure 1 that despite the presence of moderate amount of fibrosis (Fig. 1l), fatty change (steatosis) was absent (Fig. 1k) in the treated mouse. Despite the substantial increase in the necrosis, inflammation, and fibrosis scores, ADC did not differ between any of the animal groups (P > 0.191 for all; Fig. 2d). In contrast, FA of the F2 group was significantly lower than that of control (0.539 6 0.058 vs. 0.586 6 0.013; P < 0.008; Fig. 2e). Due to the relatively small numbers of animals in F1 (n ¼ 4) and F3 (n ¼ 5), all CCl4-treated mice were grouped together as a treated group (n ¼ 20). Correlations Between Histopathologic Parameters HFF (ranging 2.8–8.9% with the mean 6 SD of 4.92 6 1.32 % for all animals) was not correlated with any of the necrosis (r ¼ 0.003, P ¼ 0.986), inflammation (r ¼ 0.020,
DTI in Liver Fibrosis with Minimal Steatosis
FIG. 2. Comparison among the animal groups in HFF (a), necrosis (b), inflammation (c), ADC (d), FA (e), age (f), and weight (g). F1–F3: fibrosis scores, *statistically significant difference (a ¼ 0.008; in these multiple pair-wise group comparisons the significance level, a, was adjusted from 0.05 to 0.008 after Bonferroni correction).
P ¼ 0.906), and fibrosis (r ¼ 0.173, P ¼ 0.313) scores. Necrosis was correlated with inflammation and fibrosis (r ¼ 0.728 and 0.680; P < 0.001 for both). Inflammation was also correlated with fibrosis (r ¼ 0.754, P < 0.001). Comparison Between the Control and the Treated Groups The comparison between the control (n ¼ 16) and the treated groups (n ¼ 20) in age, weight, HFF, histopathologic parameters, and DTI parameters is summarized in Figure 3. The mean HFFs of the two groups (Fig. 3a) were both below 5.5% and did not differ from each other (5.3 6 1.5 vs. 4.6 6 1.1%; P ¼ 0.115). On the other hand, the necrosis, inflammation, and fibrosis scores were all significantly higher in the treated group with respect to control (Fig. 3b-d).
Despite the significantly elevated disease severity in the treated group, there was no difference in ADC between the two animal groups (0.711 6 0.068 103 vs. 0.718 6 0.095 103 mm2 s1; P ¼ 0.911; Fig. 3e). On the other hand, FA of the treated group was significantly lower than that of control (0.552 6 0.050 vs. 0.586 6 0.013; P ¼ 0.023; Fig. 3f). Correlations Between DTI Parameters and Histopathology ADC was not correlated with any of HFF, necrosis, inflammation, and fibrosis (r ¼ 0.190, 0.106, 0.144, and 0.089, respectively; P > 0.266 for all). FA was not correlated with necrosis and inflammation (r ¼ 0.198 and 0.141, respectively; P > 0.246 for both) but was positively correlated with HFF (r ¼ 0.418, P ¼ 0.011) and negatively correlated with fibrosis (r ¼ 0.411, P ¼ 0.012).
FIG. 3. Comparison between the control and the treated groups in HFF (a), necrosis (b), inflammation (c), fibrosis (d), ADC (e), FA (f), age (g), and weight (h) (*statistically significant difference [a ¼ 0.05]).
DISCUSSION Diffusion MRI, particularly diffusion weighted magnetic resonance imaging, is increasingly used in routine abdominal MRI protocol, where ADC has been considered a potential MR biomarker in staging liver fibrosis (1,3,5). However, recent studies have shown that the presence of fat in the hepatocytes itself (6,10,13,14) and/ or the residual fat signal resulting from imperfect fat suppression (6,12,15) can alter observed ADC in the liver. For instance, a strong negative correlation between the degree of hepatic steatosis and ADC has been reported in an ex vivo animal study (12). In a recent human study, such an inverse relationship between ADC and hepatic fat content has also been reported (13). Neither of these studies, however, investigated the direct relationship between ADC and fibrosis. Given that fatty liver is common in liver diseases, therefore, the question arises as to the performance of diffusion MRI in the assessment of liver fibrosis if the potential confounding effect of fat on ADC measurement is accounted for, which was the motivation of our current study. In addressing such an issue, the difficulty with the concomitant fatty liver in liver fibrosis is not limited to human study. For instance, CCl4 intoxication in rats is a well-established, widely used liver fibrosis model (21,22,33). However, the animal model can develop severe steatosis as the disease evolves (23,24), and therefore is not suitable for current investigation. In contrast, the response of the mice (C57BL/6) to CCl4 insult in our study is quite different from that of rats in that the hepatic fat content at the baseline was almost retained throughout the course of the evolution of liver fibrosis with the severity score up to 3. This is in line with a previous study (20) where the C57BL/6 mice treated with CCl4 induced severe fibrosis (bridging fibrosis) but with much less hepatic fat accumulation compared to those mice on a methionine-choline-deficient diet, which developed steatohepatitis in addition to fibrosis. Therefore, our choice of this specific animal model is well justified. Another murine model of hepatic fibrosis with minimal steatosis has also been reported recently where C57BL/6 mice were used in combination with a 3,5dicarbethoxy-1,4-dihydrocollidine diet (18). The sources of difficulty in addressing current issue may also include image artifact that can often be persistent in abdominal diffusion MRI particularly in combination with an EPI readout in the presence of hepatic steatosis (12). To reduce potential errors in the estimation of diffusion MRI parameters resulting from such image degradation, we have used SE. Given that the mean HFFs of the control and the treated animal groups in our study were both below 5.5%, the confounding effect of fat resulting from the fat signal (4.2–5.3 ppm) proximal to water signal (4.7 ppm) on observed ADC should also be minimal (6,15). In our study, while there was no difference in HFF between the control and the treated groups, the necrosis, inflammation, and fibrosis scores were all substantially elevated in the treated group. Nonetheless, ADC did not differ between the two animal groups. In a previous ex vivo study using excised fibrotic livers from C57BL/6
Lee and Kim
mice with minimal steatosis, lower ADC values were found in association with higher degrees of fibrosis (18). However, in that study, 8 of 15 treated mice had a fibrosis score of 3 or higher, whereas the majority of our treated mice (15 of 20 mice) had fibrosis scores of 1 or 2. In addition, the mean inflammation score of our treated mice was only 1.40. Given that inflammation can also lower ADC (3,7), the relatively low disease severity of our experimental animals may explain at least in part the lack of sensitivity of ADC to fibrosis in our study. Together with the lack of correlations between ADC and the histopathologic parameters, therefore, ADC may not sensitively detect mild-to-moderate fibrosis. Such a poor sensitivity of ADC to the severity of liver fibrosis has also been reported previously (2,4,6,9). Instead, FA was significantly lowered in the treated group with respect to control. Therefore, given the comparable HFF between the two animal groups, the elevation of the histopathologic parameters, and its negative correlation with fibrosis altogether, FA may more sensitively detect the severity of liver fibrosis than ADC. Recently, Cheung et al. (19) investigated the potential efficacy of DTI in assessing liver fibrosis in CCl4-treated rats and found decreased FA in rats with mild fibrosis at 2 weeks after CCl4 insult. While the animals in that study developed steatosis as well, and therefore may not directly be compared to our animals, the lowered FA therein is in line with our finding. As well, the higher sensitivity of FA than ADC to fibrosis was also found in the same study. However, it should be noted that, in addition to the negative correlation with fibrosis, FA is also positively correlated with HFF in our study. This means that, unlike in our current animal model, such potential sensitivity of FA to both fibrosis and HFF may not be depicted, for instance, in human liver fibrosis where severe steatosis can also develop thereby compensating for the effect of fibrosis. In line with our finding, a higher FA in fatty liver patients with respect to normal patients has been reported previously (11). In a recent DTI liver study in humans (7), the severity of liver fibrosis was not correlated with any of the ADC and FA values. However, in that study, the severity of steatosis in diseased livers was not assessed, which may explain the lack of correlation between FA and the severity of fibrosis found therein. There have been only a few DTI liver studies to date (3,7,11,19,34), and a rather wide range of DTI-derived ADC and FA values have been reported, the former of which are independent of the orientation of the gradient axis direction (25,30) and therefore may be more suitable for interlaboratory comparison than conventional, diffusion weighted magnetic resonance imaging-derived ADC values. Specifically, the baseline ADC ( 103 mm2 s1) and FA values in humans were reported to be 1.38 and 0.46, respectively, at b ¼ 1000 s mm2 (7), 2.0 and 0.3 at b ¼ 400 s mm2 (11), and a mean ADC value of 1.47 at b ¼ 500 s mm2 (3). In rats, they were reported to be ADC 0.97–1.03 103 mm2 s1 and FA 0.21– 0.26 at b ¼ 1000 s mm2 (19,34), and, in mice, we have obtained ADC ¼ 0.711 103 mm2 s1 and FA ¼ 0.586 at b ¼ 500 s mm2. Further studies are required to clarify where or not species difference is also one of the factors responsible for these variations in ADC and FA values in
DTI in Liver Fibrosis with Minimal Steatosis
addition to differences in imaging protocol. However, in a previous study, a mean hepatic ADC of 0.786 103 mm2 s1 was obtained at b ¼ 21, 301, and 601 s mm2 in normal C57BL/6 mice (12). While the data were acquired ex vivo from only three mice and the ADC values were diffusion weighted magnetic resonance imaging-derived in that study, our ADC values are more comparable to those in that study than to the others. Meanwhile, the previously reported correlation between ADC and hepatic fat content (6,10,12,13) was not found in our study. However, the mean values and dynamic ranges of hepatic fat content in these previous studies were larger than those of our animals (for instance, HFF ranging 5– 35% with the mean HFF of 13.67% in the fatty liver group in Poyraz et al.’s study (13)). Therefore, the lack of correlation between ADC and hepatic fat content in our study may likely be due to the limited dynamic range of the HFF values (2.8–8.9%) and the relatively low mean HFF value (4.92 6 1.32%) of our animals. This may, in turn, indicate that FA is more sensitive to fatty change than ADC as it is to fibrosis. In this case, our study also suggests that altered FA in the liver may also need to be carefully interpreted in the presence of fatty liver. Our study has limitations. Given the intermediate bvalue of 500 s mm2 in this study, our results may have been confounded by blood perfusion (35,36). The use of an additional, higher b-value and/or an intravoxel incoherent motion (IVIM) imaging may have allowed quantitative analysis on the extent of the perfusion effect (35,36). However, it should be noted that the use of SE instead of EPI in our study also allowed relatively higher image resolution than that which is typically obtained in a single-shot EPI readout in the liver. For instance, our voxel volume (0.019 mm3) is smaller than that (1.28 mm3) in a previous DTI rat study (19) by a factor of 67. Consequently, the effect of blood perfusion against pure water diffusion should be reduced (35). The use of isofluorane for anesthesia in our study, which is known to lower hepatic blood flow (37) may also have reduced the effect of blood perfusion to a certain degree. Fibrotic livers are known to have lower blood perfusion (2,23,36). Therefore, if the perfusion effect had influenced our diffusion MR parameters substantially, then, for instance, the ADC of the mice with higher fibrosis scores (Fig. 2) or of the treated mice group (Fig. 3) should have been lowered with respect to control. As well, given the lack of correlation between HFF and fibrosis scores, the potential sensitivity of FA to HFF found in our study should be independent of the extent of the perfusion effect. The measurement reproducibility of diffusion MRI may be influenced by ROI definitions (38). We used nonstandardized, arbitrary-shaped ROIs in an effort of including as much liver parenchyma tissue as possible while avoiding blood vessels. However, the variability in ADC and FA between the repeated measurements was less than 3% for all slices of the animals. Given that the ADC and FA values representing each animal and each animal group were obtained from the mean values over eight slices and over at least 16 animals (control vs. treated), respectively, the potential influence of using such arbitrary-shaped ROIs on our statistical outcome should be minimal. The resulting larger number of pixels
in combination with our relatively larger number of slices can also render the image analysis more robust against potential errors resulting from image degradation and liver inhomogeneity. In conclusion, at b ¼ 0 and 500 s mm2, FA may be more sensitive to mild-to-moderate liver fibrosis than ADC. In addition to ADC as demonstrated elsewhere, FA may also be sensitive to hepatic fat content, and therefore need to be carefully interpreted in the presence of fatty liver. REFERENCES 1. Taouli B, Tolia AJ, Losada M, Babb JS, Chan ES, Bannan MA, Tobias H. Diffusion-weighted MRI for quantification of liver fibrosis: preliminary experience. Am J Roentgenol 2007;189:799–806. 2. Annet L, Peeters F, Abarca-Quinones J, Leclercq I, Moulin P, Van Beers BE. Assessment of diffusion-weighted MR imaging in liver fibrosis. J Magn Reson Imaging 2007;25:122–128. 3. Taouli B, Chouli M, Martin AJ, Qayyum A, Coakley FV, Vilgrain V. Chronic hepatitis: role of diffusion-weighted imaging and diffusion tensor imaging for the diagnosis of liver fibrosis and inflammation. J Magn Reson Imaging 2008;28:89–95. 4. Sandrasegaran K, Akisik FM, Lin C, Tahir B, Rajan J, Saxena R, Aisen AM. Value of diffusion-weighted MRI for assessing liver fibrosis and cirrhosis. Am J Roentgenol 2009;193:1556–1560. 5. Taouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010;254:47–66. 6. Bulow R, Mensel B, Meffert P, Hernando D, Evert M, Kuhn JP. Diffusionweighted magnetic resonance imaging for staging liver fibrosis is less reliable in the presence of fat and iron. Eur Radiol 2013;23:1281–1287. 7. Tosun M, Inan N, Sarisoy HT, Akansel G, Gumustas S, Gurbuz Y, Demirci A. Diagnostic performance of conventional diffusion weighted imaging and diffusion tensor imaging for the liver fibrosis and inflammation. Eur J Radiol 2013;82:203–207. 8. Lewin M, Poujol-Robert A, Boelle PY, et al. Diffusion-weighted magnetic resonance imaging for the assessment of fibrosis in chronic hepatitis C. Hepatology 2007;46:658–665. 9. Boulanger Y, Amara M, Lepanto L, Beaudoin G, Nguyen BN, Allaire G, Poliquin M, Nicolet V. Diffusion-weighted MR imaging of the liver of hepatitis C patients. NMR Biomed 2003;16:132–136. 10. Wignall O, Scurr E, Collins D, Thng C-H, Koh D-M. Hepatic steatosis results in a reduction in the apparent diffusion coefficient (ADC) of liver parenchyma. In Proceedings of the 16th Annual Meeting of ISMRM, Toronto, Canada, 2008. p. 3708. 11. Lu L, Haus JM, Kirwan JP, Flask CA. Liver DTI of obese insulin resistant subjects with fatty liver disease. In Proceedings of the 18th Annual Meeting of ISMRM, Stockholm, Sweden, 2010. p. 4710. 12. Anderson SW, Soto JA, Milch HN, Ozonoff A, O’Brien M, Hamilton JA, Jara HJ. Effect of disease progression on liver apparent diffusion coefficient values in a murine model of NASH at 11.7 Tesla MRI. J Magn Reson Imaging 2011;33:882–888. 13. Poyraz AK, Onur MR, Kocakoc E, Ogur E. Diffusion-weighted MRI of fatty liver. J Magn Reson Imaging 2012;35:1108–1111. 14. Guiu B, Petit JM, Capitan V, et al. Intravoxel incoherent motion diffusion-weighted imaging in nonalcoholic fatty liver disease: a 3.0T MR study. Radiology 2012;265:96–103. 15. Hansmann J, Hernando D, Reeder SB. Fat confounds the observed apparent diffusion coefficient in patients with hepatic steatosis. Magn Reson Med 2013;69:545–552. 16. d’Assignies G, Ruel M, Khiat A, Lepanto L, Chagnon M, Kauffmann C, Tang A, Gaboury L, Boulanger Y. Noninvasive quantitation of human liver steatosis using magnetic resonance and bioassay methods. Eur Radiol 2009;19:2033–2040. 17. Lee JT, Liau J, Murphy P, Schroeder ME, Sirlin CB, Bydder M. Crosssectional investigation of correlation between hepatic steatosis and IVIM perfusion on MR imaging. Magn Reson Imaging 2012;30:572–578. 18. Anderson SW, Jara H, Ozonoff O, O’Brien M, Hamilton JA, Soto JA. Effect of disease progression on liver apparent diffusion coefficient and T2 values in a murine model of hepatic fibrosis at 11.7 Tesla MRI. J Magn Reson Imaging 2012;35:140–146. 19. Cheung JS, Fan SJ, Gao DS, Chow AM, Man K, Wu EX. Diffusion tensor imaging of liver fibrosis in an experimental model. J Magn Reson Imaging 2010;32:1141–1148.
1608 20. Baeck C, Wehr A, Karlmark KR, et al. Pharmacological inhibition of the chemokine CCL2 (MCP-1) diminishes liver macrophage infiltration and steatohepatitis in chronic hepatic injury. Gut 2012;61:416– 426. 21. Hernandez-Munoz R, Diaz-Munoz M, Lopez V, Lopez-Barrera F, Yanez L, Vidrio S, Aranda-Fraustro A, Chagoya de Sanchez V. Balance between oxidative damage and proliferative potential in an experimental rat model of CCl4-induced cirrhosis: protective role of adenosine administration. Hepatology 1997;26:1100–1110. 22. Maya-Mendoza A, Hernandez-Munoz R, Gariglio P, Aranda-Anzaldo A. Gene positional changes relative to the nuclear substructure during carbon tetrachloride-induced hepatic fibrosis in rats. J Cell Biochem 2004;93:1084–1098. 23. Kim H, Booth CJ, Pinus AB, Chen P, Lee A, Qiu M, Whitlock M, Murphy PS, Constable RT. Induced hepatic fibrosis in rats: hepatic steatosis, macromolecule content, perfusion parameters, and their correlations—preliminary MR imaging in rats. Radiology 2008;247:696–705. 24. Lee Y, Jee HJ, Noh H, et al. In vivo (1) H-MRS hepatic lipid profiling in nonalcoholic fatty liver disease: an animal study at 9.4 T. Magn Reson Med 2013;70:620–629. 25. Le Bihan D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N, Chabriat H. Diffusion tensor imaging: concepts and applications. J Magn Reson Imaging 2001;13:534–546. 26. Frahm J, Merboldt KD, Hanicke W. Localized proton spectroscopy using stimulated echoes. J Magn Reson 1987;72:502–508. 27. Yablonskiy DA, Neil JJ, Raichle ME, Ackerman JJ. Homonuclear J coupling effects in volume localized NMR spectroscopy: pitfalls and solutions. Magn Reson Med 1998;39:169–178. 28. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259–267. 29. Jiang H, van Zijl PC, Kim J, Pearlson GD, Mori S. DtiStudio: resource program for diffusion tensor computation and fiber bun-
Lee and Kim
dle tracking. Computer Methods Programs Biomed 2006;81:106– 116. Dong Q, Welsh RC, Chenevert TL, Carlos RC, Maly-Sundgren P, Gomez-Hassan DM, Mukherji SK. Clinical applications of diffusion tensor imaging. J Magn Reson Imaging 2004;19:6–18. Naressi A, Couturier C, Castang I, de Beer R, Graveron-Demilly D. Java-based graphical user interface for MRUI, a software package for quantitation of in vivo/medical magnetic resonance spectroscopy signals. Comput Biol Med 2001;31:269–286. Vanhamme L, van den Boogaart A, Van Huffel S. Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J Magn Reson 1997;129:35–43. Constandinou C, Henderson N, Iredale JP. Modeling liver fibrosis in rodents. Methods Mol Med 2005;117:237–250. Cheung JS, Fan SJ, Chow AM, Hui ES, Wu EX. In vivo DTI assessment of hepatic ischemia reperfusion injury in an experimental rat model. J Magn Reson Imaging 2009;30:890–895. Le Bihan D, Breton E, Lallemand D, Aubin ML, Vignaud J, LavalJeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988;168:497–505. Luciani A, Vignaud A, Cavet M, Nhieu JT, Mallat A, Ruel L, Laurent A, Deux JF, Brugieres P, Rahmouni A. Liver cirrhosis: intravoxel incoherent motion MR imaging—pilot study. Radiology 2008;249: 891–899. Frink EJ Jr, Morgan SE, Coetzee A, Conzen PF, Brown BR Jr. The effects of sevoflurane, halothane, enflurane, and isoflurane on hepatic blood flow and oxygenation in chronically instrumented greyhound dogs. Anesthesiology 1992;76:85–90. Colagrande S, Pasquinelli F, Mazzoni LN, Belli G, Virgili G. MR-diffusion weighted imaging of healthy liver parenchyma: repeatability and reproducibility of apparent diffusion coefficient measurement. J Magn Reson Imaging 2010;31:912–920.