To characterize changes in diffusion properties of liver using diffusion tensor imaging (DTI) in an experimental model of liver fibrosis.
To characterize changes in diffusion properties of liver using diffusion tensor imaging (DTI) in an experimental model of liver fibrosis.
Liver fibrosis was induced in Sprague–Dawley rats (n = 12) by repetitive dosing of carbon tetrachloride (CCl4). The animals were examined with a respiratory-gated single-shot spin-echo echo-planar DTI protocol at 7 T before, 2 weeks after, and 4 weeks after CCl4 insult. Apparent diffusion coefficient (ADC), directional diffusivities (ADC// and ADC⟂), and fractional anisotropy (FA) were measured. Liver histology was performed with hematoxylin-eosin staining and Masson's trichrome staining.
Significant decrease (P < 0.01) in ADC was found at 2 weeks (0.86 ± 0.09 × 10−3 mm2/s) and 4 weeks (0.74 ± 0.09 × 10−3 mm2/s) following CCl4 insult, as compared with that before insult (0.97 ± 0.08 × 10−3 mm2/s). Meanwhile, FA at 2 weeks (0.18 ± 0.03) after CCl4 insult was significantly lower (P < 0.01) than that before insult (0.26 ± 0.05), and subsequently normalized at 4 weeks (0.26 ± 0.07) after the insult. Histology showed collagen deposition, presence of intracellular fat vacuoles, and cell necrosis/apoptosis in livers with CCl4 insult.
DTI detected the progressive changes in water diffusivities and diffusion anisotropy of liver tissue in this liver fibrosis model. ADC and FA are potentially valuable in detecting liver fibrosis at early stages and monitoring its progression. Future human studies are warranted to further verify the applicability of DTI in characterizing liver fibrosis and to determine its role in clinical settings. J. Magn. Reson. Imaging 2010;32:1141–1148. © 2010 Wiley-Liss, Inc.
LIVER FIBROSIS is a typical complication of chronic liver diseases resulting in cirrhosis and increased risk of hepatocellular carcinoma (1–5). More than 400 and 170 million individuals are chronically infected with hepatitis B and hepatitis C virus, respectively, worldwide (6). Early diagnosis and characterization of liver fibrosis could facilitate early interventions and thus prevent its progression to cirrhosis (7–9). However, the efficacy of the current noninvasive techniques in assessing liver fibrosis such as routine liver function tests, serological tests of specific serum makers, and liver stiffness measurement using ultrasound transient elastography has yet to be established (6, 10–13). At present, percutaneous liver biopsy has been considered the gold standard, yet it is highly invasive. Liver biopsy is also associated with potential risks of complications and repeated biopsies are not practicable clinically, thus its utility as a tool for longitudinal monitoring has been limited. In addition, it is prone to sampling error and interobserver variation (14, 15), leading to erroneous staging. Therefore, a noninvasive magnetic resonance imaging (MRI) technique for detecting liver fibrosis at early stages and monitoring the disease progression or regression in response to treatment is highly desirable.
Morphologic analysis using conventional MRI for the assessment of liver fibrosis and cirrhosis has been subjected to interobserver variability and limited in sensitivity and specificity (16–18). Several novel MRI techniques including contrast-enhanced MRI, MR elastography, and MR diffusion imaging have been proposed as noninvasive alternatives to characterize liver fibrosis and cirrhosis (18, 19). Double-contrast MRI using both gadolinium chelates and superparamagnetic iron oxides (SPIOs) was suggested to provide synergistic effects in visualizing liver fibrosis directly based on the hepatic texture alterations. Specifically, SPIOs preferentially accumulated and darkened nonfibrous liver tissue, while gadolinium chelates caused signal enhancement in fibrous tissues which appeared as bright reticulations (20, 21). MR elastography has been recently shown to be sensitive in assessing liver fibrosis by measuring the mechanical properties of liver tissue, in particular, tissue elasticity. Liver stiffness measured by MR elastography was found to increase systematically with the stage of liver fibrosis (22, 23). However, more studies remain to be performed to fully evaluate the clinical utility of these two MRI techniques for staging liver fibrosis.
MR diffusion imaging in liver, which can be easily incorporated into routine MRI examination on most clinical scanners, is a sensitive tool for characterizing the microscopic motion of water molecules (24–26). Molecular diffusion of water molecules arises from their random motion at a microscopic level. Diffusion-weighted imaging (DWI) and diffusion tensor imaging (DTI) have been widely used to characterize normal and diseased tissues the in central nervous system since its introduction (24–27). With the advent of the single-shot echo-planar imaging (EPI) technique in combination with respiratory triggering or breath-holding (28–31), MR diffusion imaging has become possible in abdominal organs regardless of the effect of gross physiologic motion. Apparent diffusion coefficient (ADC) measured by DWI has been used to characterize focal hepatic lesions (32–36) and diffuse liver diseases including liver fibrosis (37–46). Liver fibrosis is a nonspecific response to chronic liver disease which leads to excess synthesis of extracellular matrix (ECM), especially collagen fibers, in which the protons are less abundant and are tightly bound (47, 48). Several studies hypothesized that molecular water diffusion in fibrotic liver would be restricted by the presence of collagen fibers in the distorted lobular structure and their results showed that ADC values decreased in fibrotic and cirrhotic liver as compared with normal liver (38–45).
DTI has the potential to provide both functional (ie, water diffusion in tissue) and microstructural information in liver tissue by means of water diffusivity and diffusion anisotropy quantitation, and thus may contribute to better characterization of liver fibrosis. The aim of this study was to characterize changes in diffusion properties of liver using DTI in an experimental model of liver fibrosis.
All animal experiments were approved by the institutional animal ethics committee. Liver fibrosis was induced in male adult Sprague–Dawley rats (220–260 g; n = 12) by subcutaneous injection of 1:1 volume mixture of carbon tetrachloride (CCl4) in olive oil at a dose of 0.2 mL / 100 g of body weight twice a week for 4 weeks (49–51). Intermittent administration of CCl4 has been widely used to experimentally induce liver fibrosis in rodents by evoking a marked infiltration of inflammatory cells, thus mimicking the changes in chronic viral hepatitis-associated fibrosis (49, 52). The twice-weekly dosing can induce early stages of liver fibrosis and established fibrosis after 2 and 4 weeks of CCl4 administration, respectively, in rodents (49, 51). This well-controlled CCl4-induced liver fibrosis model allows the study of a homogeneous population of liver fibrosis. Prior to MRI experiments, animals were fasted to reduce motion artifact due to peristalsis (53). MRI was performed in animals at 1 day before, 2 weeks after, and 4 weeks following CCl4 administration. Note that animals were imaged at about 48 hours after the last CCl4 administration to avoid any acute toxic or inflammatory responses of CCl4 on MRI measurements (54). The overall schedule of the animal experiment is shown in Fig. 1.
All MRI experiments were performed on a 7 T MRI scanner with a maximum gradient of 360 mT/m (70/16 PharmaScan, Bruker Biospin, Germany), using a 60 mm quadrature resonator for both radiofrequency (RF) transmission and receiving. During imaging, each animal was anesthetized with isoflurane/air using 1.0%–1.5% for maintenance via a nose cone. Body temperature was maintained at ≈36.5°C by circulating warm water in a heating pad. A respiratory belt was placed around the abdomen to monitor respiration and synchronize MR acquisition. T1-weighted scout scans were acquired in three orthogonal directions to ensure similar slice localization for DTI in different animals at different timepoints. DTI was performed in one axial slice covering a large portion of liver while avoiding inclusion of the lung. The respiratory-gated single-shot spin-echo EPI (SE-EPI) DTI protocol was employed using repetition time (TR) = 2 respiratory cycles (≈2.0–2.5 sec), echo time (TE) = 32 msec, duration of gradient pulse/diffusion time (δ/Δ) = 2.6/20 msec, two b-values (0 and 1000 s/mm2), six diffusion gradient directions, field of view (FOV) = 5.12 × 5.12 cm2, slice thickness = 2 mm, acquisition matrix = 64 × 64, voxel size = 0.8 × 0.8 × 2 mm3, sampling bandwidth (BW) = 221 kHz, number of signal averages (NEX) = 10, and scan time of ≈3 minutes (30, 31). Note that a high b-value (1000 s/mm2) was used to decrease the effect of blood perfusion on diffusion measurements (30, 32, 55). The DTI acquisition was repeated twice. A T2-weighted image at the same slice location was acquired with a 2D rapid acquisition with relaxation enhancement (RARE) protocol using TR = 2000 msec, effective TE = 40 msec, FOV = 5.12 × 5.12 cm2, slice thickness = 2.0 mm, acquisition matrix = 128 × 128, RARE factor = 8, and with respiratory gating. Note that a relatively small matrix size was chosen in single-shot SE-EPI DTI to maintain sufficient signal-to-noise ratio (SNR) and acceptable level of EPI-related artifacts such as geometric distortion and N/2 ghosting.
Image analysis was performed on a blind basis with respect to the status of the animals. Diffusion-weighted EPI images were coregistered with the nondiffusion-weighted (B0) EPI image using Automated Image Registration (AIR 5.2.5) (56) for correcting any misregistration caused by the body motions during DTI and the gradient eddy current related image distortions. ADC, axial diffusivity (ADC//), radial diffusivity (ADC⟂), fractional anisotropy (FA), and color-coded FA (CFA) direction maps were generated using DTIStudio (v. 2.4; JHU, Baltimore, MD) (57). In brief, three principal diffusivities (λ1, λ2, and λ3) were derived from diagonalizating the 3 × 3 diffusion tensor matrix. Afterwards, ADC//, ADC⟂, ADC, and FA were computed as follows (25, 26, 30, 31):
Note that water diffusivities parallel and perpendicular to the principal diffusion direction are represented by ADC// and ADC⟂, respectively. Moreover, ADC is a rotationally invariant measure of the overall water diffusivity, while FA is a measure of diffusion directionality. A region of interest (ROI) was first defined in the axial anatomical scout image acquired before CCl4 insult by encompassing a large homogeneous region in liver parenchyma while excluding major blood vessels. It was used for ADC, λ//, λ⟂, and FA measurements. For measurements at subsequent timepoints the same ROI mask was placed onto the anatomical scout image with slight manual adjustments to account for misregistration. The measurements of repeated DTI acquisitions were averaged for each animal and one-way analysis of variance (ANOVA) with Tukey's multiple comparison test was employed to compare the ADC, ADC//, ADC⟂, and FA measurements among different timepoints in animals. Results were expressed as mean ± standard deviation (SD). A P-value of less than 0.05 was considered statistically significant.
After the MRI scan following 2 weeks of CCl4 administration, 4 out of 12 animals were sacrificed for histological evaluation. Furthermore, four out of the remaining eight animals were sacrificed after MRI following 4 weeks of CCl4 insult as shown in Fig. 1. One additional normal animal was sacrificed as a control. Liver specimens were fixed in formalin, embedded in paraffin, sectioned, and examined by light microscopy after standard hematoxylin-eosin (H&E) staining and Masson's trichrome staining (58–60). Degree of liver fibrosis in the samples was evaluated semiquantitatively according to the METAVIR classification using a five-point scale with F0, F1, F2, F3, and F4 for no fibrosis, portal fibrosis without septa, portal fibrosis with few septa, numerous septa without cirrhosis, and cirrhosis, respectively (61).
Figure 2 shows the representative B0 EPI images, T2-weighted images, ADC maps, FA maps, and CFA direction maps of liver from one animal before, 2 weeks after, and 4 weeks after CCl4 insult. Typical ROIs used for ADC, ADC//, ADC⟂, and FA are illustrated in the B0 EPI images with size of 95 ± 27 mm2 in all animals at different timepoints. Note that FA in the liver was seen to be relatively inhomogeneous without any directional dominance, while the dominant diffusion direction in the spinal cord was consistently observed to be along the cranial–caudal direction (blue in the CFA maps) as expected. ADC, FA, ADC//, and ADC⟂ values at different timepoints of CCl4 insult are shown in Fig. 3. A significant decrease (P < 0.01) in ADC was found at 2 weeks (0.86 ± 0.09 × 10−3 mm2/s) and 4 weeks (0.74 ± 0.09 × 10−3 mm2/s) following CCl4 insult, as compared with that before insult (0.97 ± 0.08 × 10−3 mm2/s). Meanwhile, FA at 2 weeks (0.18 ± 0.03) after CCl4 insult was significantly lower (P < 0.01) than that before (0.26 ± 0.05) and 4 weeks after (0.26 ± 0.07) the insult. FA at 4 weeks after CCl4 insult was not significantly different from that before insult. Moreover, ADC// at 2 weeks (1.01 ± 0.09 × 10−3 mm2/s) and 4 weeks (0.95 ± 0.09 × 10−3 mm2/s) after insult was found to be significantly lower (P < 0.01) than that before insult (1.24 ± 0.15 × 10−3 mm2/s), whereas ADC⟂ at 4 weeks (0.64 ± 0.09 × 10−3 mm2/s) after insult was significantly lower (P < 0.01) than that before (0.84 ± 0.06 × 10−3 mm2/s) and 2 weeks after (0.78 ± 0.09 × 10−3 mm2/s) insult.
Figure 4 shows the typical H&E and Masson's trichrome staining of normal liver and livers at 2 weeks and 4 weeks after CCl4 insult. Collagen deposition was stained blue by Masson's trichrome staining in fibrotic livers. Compared with normal liver (Fig. 4a), collagen deposition and intracellular fat vacuoles were consistently observed in livers with CCl4 insult (Fig. 4b,c). Cell necrosis/apoptosis was evident in liver with 2-week CCl4 insult (Fig. 4b), while collagen deposition was more pronounced in liver with 4-week CCl4 insult (Fig. 4c). Similar histological findings were observed in all liver samples collected, and they were largely consistent with those from the earlier studies of CCl4-induced liver fibrosis in rodent models (54). Specifically, the liver specimens collected at 2 weeks after fibrosis induction showed scattered collagen deposition in the sinusoids of the pericentral lobular area without septa formation (Fig. 4b), representing mild fibrosis. The specimens collected at 4 weeks exhibited clear septa formation with dense collagen deposition forming portal–portal and portal–central bridges (Fig. 4c), indicating advanced fibrosis. Based on the semiquantitative METAVIR staging, week-2 liver specimens were classified as F1 (portal fibrosis without septa, n = 2) and F2 (portal fibrosis with few septa, n = 2). Week-4 specimens were staged as F2 (n = 1) and F3 (numerous septa without cirrhosis, n = 3). Normal liver yielded F0 (no fibrosis, n = 1) as expected.
Liver fibrosis is an important factor of hepatic dysfunction, yet its progression can be reversed by suppressing the underlying cause (7–9). The standard diagnosis and staging of liver fibrosis currently rely on histological analysis through liver biopsy. However, it is considerably invasive and often associated with potential complications, sampling errors, and interobserver variability (14, 15). Accordingly, the development of alternative noninvasive techniques to characterize liver fibrosis is urgently needed. MR diffusion imaging, a routine MRI protocol in studying brains in both humans and animal models (24–26), is a sensitive and noninvasive technique for characterizing the random microscopic motion of water molecules. Thus, DTI may provide information on the microstructural environments in liver tissue and reveal the pathological alterations in tissue microstructure during liver fibrogenesis. In this study we characterized the changes in diffusion properties of liver using DTI in a well-controlled experimental model of liver fibrosis. Note that the CCl4-intoxication fibrosis model employed in the current study is the most commonly recognized and employed experimental model to investigate the cellular and molecular mediators involved in fibrosis that are highly relevant to those in human (49).
The histological observations revealed the collagen deposition and thus confirmed liver fibrogenesis in the animals studied. Despite the potential water diffusivity increase associated with cell necrosis/apoptosis, ADC was observed to decrease gradually with CCl4 insult, likely due to the increased extracellular collagen deposition and increased intracellular fat droplets during the progression of liver fibrosis (Fig. 4b,c). In addition, the ADC decrease could also be associated with the decreased blood perfusion as suggested by several other studies (39, 43, 44) despite the relatively high b-value used. FA decrease at 2 weeks after CCl4 insult resulted from the significant decrease of ADC// (P < 0.01). This was likely due to the prominent cell necrosis/apoptosis that occurred (Fig. 4b), perturbing water diffusion along the radially oriented cellular structures in hepatic plates. The subsequent FA increase at 4 weeks after CCl4 insult arose from the significant decrease of ADC⟂ (P < 0.01) thereafter. This might be caused by the pronounced extracellular collagen deposition (Fig. 4c), leading to decreased water diffusivity in the more isotropic extracellular compartment. Note that the reduction and subsequent normalization of FA were the direct consequence of the distinct changes of ADC// and ADC⟂ at 2 and 4 weeks after fibrosis induction (as shown in Fig. 3), respectively. It is possible that such differential changes in ADC// and ADC⟂ could arise from a number of concomitant alterations in the tissue microenvironment, including collagen deposition, fatty infiltration, hepatitis, cell necrosis/apoptosis, inflammatory cell infiltration, and fibroblast proliferation with different degrees, leading to the observed FA changes. Consistent with our results, several groups have reported that ADC values were lower in fibrotic and cirrhotic liver as compared with normal liver in both human and animal studies (38–45) despite of the different b-values used. To the best of our knowledge, there has been no study investigating the diffusion anisotropy (eg, FA values) in liver fibrosis, although the DTI protocol has been employed (45). Nonetheless, the degree of diffusion anisotropy may change with the disturbed microstructure in fibrotic liver, as observed in the current study. It is worth noting that the percentage change in FA at 2 weeks after CCl4 insult (32 ± 11%) was higher than that in ADC (12 ± 3%) at 2 weeks after insult, indicating that FA could provide higher sensitivity in detecting early liver fibrosis.
The combined effects of water diffusion and capillary blood perfusion in tissues have been reported to be measured by MR diffusion imaging (55). ADC could be overestimated at low b-values due to the increased signal attenuation from intravoxel spin dephasing caused by the pseudorandom blood flow in presence of diffusion gradient. On the other hand, the true diffusion measurement would be affected less at high b-values where the diffusion attenuation would be mainly a result of molecular water diffusion because the blood signal will be mostly suppressed by the large diffusion gradients (32, 55). In this study, the ADC value measured in normal rat livers using a b-value of 1000 s/mm2 was 0.97 ± 0.08 × 10−3 mm2/s, which was lower than those reported previously (1.54 ± 0.29 × 10−3 mm2/s) measured at a b-value of 500 s/mm2 (39). This discrepancy likely resulted from the lesser contribution of the pseudorandom blood perfusion to the apparent diffusion measurement in the current study due to the higher b-value used. Note that the actual imaging sequences and imaging parameters, such as the effects of blood flow under a specific sequence setting and voxel size, could also influence the extent of perfusion contribution to the apparent diffusion measurement. In addition, several studies suggested that microvascular perfusion may be important in assessing liver fibrosis (39, 43). In this regard, an intravoxel incoherent motion (IVIM) model (55) has been proposed to separate the effect of blood perfusion from water diffusion (44). Further studies using IVIM analysis are needed to assess the relative contribution of these two effects on ADC measurement in fibrotic liver.
Several limitations exist in the current study. First, comprehensive histopathological correlation was not performed to examine and validate the mechanisms underlying the DTI index changes observed in this study. In fact, as in DTI of neural tissue, the direct correlation between DTI indices and specific tissue morphological characteristics can be complex and problematic, if not impossible, because the water diffusion process in vivo is affected by numerous and complex determinants, including cellular microstructures, membrane permeability or water exchange, and possibly other biophysical properties associated with different water populations. Nonetheless, such validation or correlation study is highly desired in future investigation. Second, the development and progression of liver fibrosis in the CCl4-intoxication fibrosis model differs from those in human due to different etiology of the disease. In particular, the repetitive dosing of CCl4 in animals can induce fibrosis development within weeks, while the development of fibrosis in patients with liver diseases is more gradual and may take several months or years, depending on the underlying causes. Despite different etiology, the underlying pathophysiologic processes of fibrogenesis (eg, inflammatory cell infiltration, cell necrosis, fibroblast proliferation, pronounced ECM/collagen deposition, and potential progression to cirrhosis) are largely similar (62, 63). In addition, the current finding of ADC reduction in the experimental fibrosis model is in accordance with those reported in patients with liver fibrosis (38, 41, 42, 45, 46), suggesting that the liver tissue microstructural alterations may act on the diffusion parameters in a similar manner. Third, since a low b-value of 0 s/mm2 was used along with a high b-value in the current DTI acquisition, the ADC quantitation could be affected by the pseudodiffusion effect of blood perfusion (30, 55, 64). Acquiring more DWIs with multiple and higher b-values can improve the accuracy of the true ADC measurements at the cost of substantially longer scan time. Furthermore, the spatial resolution was relatively low in the single-shot EPI-DTI acquisition. It can be improved by increasing acquisition matrix via shimming improvement or/and susceptibility reduction. Lastly, the findings from the current study were preliminary, given the limited sample sizes (n = 12). Statistical power could be further increased by larger sample sizes in future studies.
With recent advances in fast imaging techniques, especially single-shot EPI, DWI and DTI have been increasingly used in imaging abdominal organs including liver at clinical field strength of 1.5 and 3 T (33–35, 65, 66). At present, DWI and DTI are implemented on most clinical scanners and can be readily incorporated into routine liver examinations to study liver fibrosis. In contrast, the MR elastography that has been recently proposed for liver fibrosis assessment requires additional hardware and software, consequently limiting its availability in most clinical settings (22, 23). Our results suggest that the microstructural alterations in fibrotic liver may vary the degree of water molecules diffusion anisotropy. As such, ADC values measured by conventional DWI can depend on the direction of diffusion gradient applied, and thus are less reproducible and more difficult for comparative studies. DTI should provide better ADC quantitation because DTI-derived ADC is rotationally invariant.
In conclusion, the experimental results in this study demonstrated that DTI could detect longitudinal changes in diffusion properties of liver in an experimental model of liver fibrosis. Changes in directional diffusivities (ADC// and ADC⟂) and FA may reveal functional (ie, water diffusivity) and microstructural changes, respectively, during the progression of liver fibrosis. Therefore, DTI is potentially valuable in detecting and characterizing liver fibrosis at early stages and monitoring its progression and related interventions in a noninvasive manner. Future human studies are warranted to further verify the applicability of DTI in characterizing liver fibrosis and to determine its role in clinical settings.