Apparent diffusion coefficient measurements with diffusion-weighted magnetic resonance imaging for evaluation of hepatic fibrosis

Authors

  • Miwa Koinuma MD,

    1. Department of Radiology, School of Medicine, Tokyo Medical and Dental University, Tokyo, Japan
    Search for more papers by this author
  • Isamu Ohashi MD,

    Corresponding author
    1. Department of Radiology, School of Medicine, Tokyo Medical and Dental University, Tokyo, Japan
    • Department of Radiology, School of Medicine, Tokyo Medical and Dental University, 1-5-45, Yushima, Bunkyo-ku, Tokyo 113-8519, Japan
    Search for more papers by this author
  • Kaoru Hanafusa MD,

    1. Department of Radiology, School of Medicine, Tokyo Medical and Dental University, Tokyo, Japan
    2. Department of Radiology, Toride Kyodo General Hospital, Ibaraki, Japan
    Search for more papers by this author
  • Hitoshi Shibuya MD

    1. Department of Radiology, School of Medicine, Tokyo Medical and Dental University, Tokyo, Japan
    Search for more papers by this author

Abstract

Purpose

To evaluate the use of apparent diffusion coefficient (ADC) measurements based on diffusion-weighted MRI (DWI) to assess stage of liver disease.

Materials and Methods

A total of 31 patients who underwent both a liver biopsy and DWI and 132 patients who only underwent DWI were enrolled. Biopsy specimens were scored for fibrosis and necroinflammation according to the Knodell histology activity index (HAI). The 31 patients consisted of 21 patients with chronic hepatitis and 10 with cirrhosis (Child-Pugh stage A in nine and stage B in one), and the 132 patients consisted of 56 patients with cirrhosis (Child-Pugh stage A in 41, stage B in 10, and stage C in five), 42 with chronic hepatitis, and 34 with normal liver function. The ADCs in the liver parenchyma were measured using DWI with relatively low b factors (b = 0.01 and 128.01 seconds/mm2) and were compared among the HAI scores and among patients with cirrhosis, chronic hepatitis, and normal liver function.

Results

The ADCs decreased as the fibrosis score in the HAI increased, and the correlation was statistically significant (P < 0.0001). No relationship between the ADCs and the necroinflammation scores in the HAI was found. The ADCs decreased as the stage of liver disease progressed or as the Child-Pugh stage progressed, and these relationships were statistically significant (P < 0.0001).

Conclusion

ADC measurements are potentially useful for the evaluation of fibrosis staging in the liver. J. Magn. Reson. Imaging 2005;22:80–85. © 2005 Wiley-Liss, Inc.

LIVER HISTOLOGY has been considered the gold standard for assessing hepatic fibrosis. In 1981, Knodell et al (1) introduced a semiquantitative and reproducible scoring system for liver biopsies, the Knodell histology activity index (HAI) score. A good correlation was seen between conventional histological descriptions for liver biopsy specimens and the HAI scores. The HAI scoring system has been widely used to evaluate hepatic necroinflammation and fibrosis. However, liver biopsy is associated with major complications in 0.3% of patients and with mortality in 0.018% (2). No other clinical symptoms, routine biochemical or hematologic blood tests, or serum markers can be used to accurately predict hepatic fibrosis (3). Therefore, an urgent need exists for a simple, noninvasive means of predicting the stage of fibrosis.

MRI is currently the best imaging method for the in vivo quantification of the combined effects of capillary perfusion and diffusion using apparent diffusion coefficient (ADC) measurement (4, 5). With the advent of the echoplanar imaging technique, diffusion-weighted MRI (DWI) of the abdomen has become possible. Several reports on the results of DWI for normal and cirrhotic livers have been made and these results suggest a lower ADC value in cirrhotic livers, compared with normal livers (6–11). This phenomenon is thought to reflect the restriction of water diffusion in fibrotic liver. However, only a few reports have compared ADC values with histologic fibrosis grades (12).

We performed echoplanar DWI in patients with various stages of chronic liver disease, such as chronic hepatitis or cirrhosis, and the measured ADC values were compared with the HAI scores for liver biopsy specimens. The purpose of the present study was to evaluate the use of noninvasive ADC measurements obtained using DWI for the assessment of liver disease, especially of hepatic fibrosis or fibrosis staging.

MATERIALS AND METHODS

Patients

First, we retrospectively evaluated 31 consecutive patients who underwent both MRI of the upper abdomen (including an echoplanar DWI) and a liver biopsy between 1994 and 2001. All patients underwent a liver biopsy to enable a histopathological evaluation and subsequent MR imaging for further clinical evaluations (mean interval two months) because of liver dysfunction related to viral hepatitis C (N = 26), B (N = 3), or other causes (N = 2). A total of 21 patients were clinically diagnosed as having chronic hepatitis, and 10 patients were diagnosed as having cirrhosis (nine patients as Child-Pugh stage A and one as stage B). The 31 patients (24 men and seven women) ranged in age from 33 to 75 years (mean = 50.4 years). All biopsies were reviewed by experienced hepatopathologists, and the histology was scored according to the Knodell HAI. The HAI system consists of the evaluation of two histopathological categories, necroinflammation and fibrosis. Furthermore, necroinflammation includes three subcategories: periportal necrosis and inflammation, scored from 0 to 10; intralobular necrosis and inflammation, scored from 0 to 4; and portal inflammation, scored from 0 to 4. Fibrosis is scored as 0, 1, 3, or 4, with 1 indicating portal fibrosis only, 3 indicating bridging fibrosis, and 4 indicating cirrhosis. The HAI score is the combined scores for necrosis, inflammation, and fibrosis, while the overall HAI scores can also be broken into individual components of necrosis, inflammation, and fibrosis to yield additional information (1).

Second, we retrospectively evaluated 143 consecutive patients who underwent MRI of the upper abdomen, including echoplanar DWI, but who did not undergo a liver biopsy between January and October 2000. Six patients were excluded because of multiple liver tumors, three patients because of poor image quality, and two patients because of marked iron deposition within the liver parenchyma. The remaining 132 patients (77 men and 55 women) ranged in age from 30 to 85 years (mean = 62.6 years). Among them, 56 patients had clinically diagnosed cirrhosis related to viral hepatitis C (N = 45), B (N = 5), or other causes (N = 6), and 42 patients had clinically diagnosed chronic hepatitis related to viral hepatitis C (N = 28), B (N = 7), or other causes (N = 7). Of the 56 patients with cirrhosis, 41 patients were graded as Child-Pugh stage A, 10 as stage B, and five as stage C. These patients underwent MRI as part of additional clinical evaluations of the liver. The remaining 34 patients had normal liver function and underwent MRI of the upper abdomen to evaluate the presence of gallstones (N = 3), hepatic hemangiomas (N = 11), cysts (N = 3), hepatic metastases from gastric cancers (N = 5), colorectal cancers (N = 9), uterine cancers (N = 2), and ovarian cancer (N = 1).

The clinical diagnosis of cirrhosis was assigned based on the combination of physical findings, such as jaundice, splenomegaly, or ascites; routine biochemical and hematologic blood tests, such as a decreased platelet count, an increased ratio of aspartate to alanine aminotransferase (AST/ALT), or a prolonged prothrombin time; and radiologic imaging studies, such as an enlarged spleen, small nodular liver, ascites, or the presence of a portosystemic shunt. Moreover, laboratory data, such as the albumin and bilirubin level and the prothrombin time, and clinical data, such as the presence or absence of ascites and hepatic encephalopathy, were collected for the patients with cirrhosis, and the Child-Pugh score was determined to evaluate the severity of the cirrhosis. A clinical diagnosis of chronic hepatitis was made based on the presence of chronic liver dysfunction with no signs suggestive of cirrhosis.

The liver biopsy and MR examination, including DWI, were initiated after the patients had been fully informed and had provided their consent. Informed consent was also obtained from all patients regarding the use of clinical data, imaging data, and specimens for the purpose of this study.

MR Protocols

We used a 1.5-T, commercially available superconducting MRI unit (Magnetom Vision; Siemens, Erlangen, Germany) with a circular polarization (CP)-body array coil for all imaging procedures. Following routine T1- and T2-weighted imaging, a series of DWIs were obtained in all patients while the patients held their breath for five seconds during expiration to avoid respiratory motion. DWI was acquired in the axial plane by combining a single-shot spin-echo type echoplanar imaging pulse sequence and additional motion probing gradient (MPG) pulses. In this sequence, 90° and 180° radio frequency (RF) pulse series were applied, and two MPG pulses were additionally applied before and after the 180° RF pulse along the z-axis. For the gradient factor b, b1 = 0.01 and b2 = 128.01 seconds/mm2 were available. The other imaging parameters of the DWI were an 81-msec TE, a 2080-Hz/pixel bandwidth, 350-mm field of view, 128 × 128 matrix, 8-mm section thickness, 2-mm interscan gap, and one signal acquisition. The data were collected using an echoplanar readout. Fat suppression was added to the DWI by placing the frequency-selective RF pulse before the pulse sequence to avoid severe chemical shift artifacts.

Each DWI series consisted of 10 to 20 sequential sections covering the whole liver. The image acquisition was repeated while the patient was in the same body position for the different MPG pulses (b1 and b2), because, in the MR unit used in this study, DWIs of different b factors were not acquired in one series, and two DWIs with different b values were obtained for each section. The total DWI time was about one minute.

Measurement of the ADC

The ADC value was calculated according to the following formula: ADC = −(lnSb2 − lnSb1)/(b2b1), where ln is the natural log, and Sb1 and Sb2 are the signal intensities in the region of interest (ROI) placed on sections corresponding to the two different b values (b1 and b2). For each series with different b values, the ROIs were positioned in the parenchyma of three consecutive mid-sections of the liver as large as possible to avoid focal lesions, major vascular structures, and artifacts, such as chemical shifts and magnetic susceptibility. The ROIs were manually and carefully positioned to be in the same region between two corresponding DWIs of different b factors. Three signal intensity values for each series were then measured. Using the above-mentioned formula, the ADC values were calculated and averaged for each patient.

Data Analysis

In the 31 patients who underwent a liver biopsy, the ADC values were correlated with each HAI category (the fibrosis score, the periportal necrosis and inflammation score, the intralobular necrosis and inflammation score, the portal inflammation score, and the total necroinflammation score) using a nonparametric Spearman's correlation coefficient by rank test.

In the 132 patients who did not undergo a liver biopsy, the ADC values were compared among patients with normal liver function, chronic hepatitis, and cirrhosis using an analysis of variance (ANOVA). The Fisher's protected least significant difference (PLSD) was used as a multiple comparison test. Furthermore, in the 56 patients with cirrhosis, the ADC values were compared among the Child-Pugh stages using a nonparametric Kruskal-Wallis test.

Differences in the ADC values were considered to be statistically significant when P < 0.05. The statistical analysis was performed using StatView software (USA).

RESULTS

The HAI scores in the patients who underwent a liver biopsy are shown for every category in Table 1. The ADC values according to fibrosis score are shown in scattered plots (Fig. 1). The ADC decreased as the fibrosis score increased, and the correlation was statistically significant (ρ = 0.798; P < 0.0001). No relationship with the ADC values was seen for any of the other HAI categories, such as periportal necrosis and inflammation, intralobular necrosis and inflammation, and portal inflammation.

Table 1. HAI Category and Score in 31 Patients Who Underwent Liver Biopsy
ScoreNecroinflammationFibrosis
Periportal necrosis and inflammationIntralobular necrosis and inflammationPortal inflammation
  1. HAI = histology activity index by Knodell.

02112
1141956
2
312112120
43043
Figure 1.

Scatterplots showing the ADCs according to fibrosis score in the HAI in 31 patients who underwent a liver biopsy. ADC decreases as fibrosis score increases, and the correlation is statistically significant (ρ = 0.798; P < 0.0001).

The ADCs in the 132 patients with clinically diagnosed normal liver function, chronic hepatitis, and cirrhosis are shown in Table 2 and in scattered plots (Fig. 2). The ADC values decreased as stage of liver disease progressed, and the difference among ADC values was statistically significant (P < 0.0001). Furthermore, statistically significant differences in the ADCs were seen between each pairing of patient groups. An overlap was observed in the ADCs of patients with normal liver function and chronic hepatitis and patients with chronic hepatitis and cirrhosis. However, no overlap was observed in the ADCs of patients with normal liver function and cirrhosis.

Table 2. ADC Values in Clinically Diagnosed 132 Patients According to Liver Disease
Liver diseaseNo. of patientsADC ± SD (×10−3 mm2/second)P
  1. ADC = apparent diffusion coefficient.

Normal343.45 ± 0.33 
Chronic hepatitis422.45 ± 0.25<0.0001
Cirrhosis561.98 ± 0.32 
Figure 2.

Scatterplots showing the ADCs according to the stage of liver disease in clinically diagnosed 132 patients. ADC decreases as liver function fails, and the relationship is statistically significant (P < 0.0001).

The ADCs in the 56 patients with cirrhosis are shown according to the Child-Pugh score of the patients in Table 3 and in scattered plots (Fig. 3). The ADC decreased as the Child-Pugh stage progressed, and the difference among the ADCs was statistically significant (P < 0.0001).

Table 3. ADC Values in 56 Patients With Cirrhosis According to Child-Pugh Score
Child-Pugh stageNo. of patientsADC ± SD (×10−3 mm2/second)P
  1. ADC = apparent diffusion coefficient.

A412.08 ± 0.28 
B101.83 ± 0.20<0.0001
C51.43 ± 0.14 
Figure 3.

Scatterplots showing the ADCs according to Child-Pugh stage in 56 patients with clinically diagnosed cirrhosis. ADC decreases as Child-Pugh stage progresses, and the relationship is statistically significant (P < 0.0001).

DISCUSSION

The typical histological features of chronic hepatitis are variable degrees of hepatocellular necrosis and inflammation (referred to as the activity or grade of disease) and fibrosis (referred to as the stage of disease) (13). While the activity of the hepatic disease can fluctuate, worsening and improving over time, the stage of fibrosis is believed to be progressive and largely irreversible. The progression of fibrosis ultimately leads to the architectural distortion of the liver and cirrhosis. For these reasons, the rate of fibrosis progression is the defining feature of the natural history of chronic hepatitis (13). However, clinical examination is unreliable for differentiating the stages of fibrosis. Among the routine laboratory tests, a decreased platelet count, increased AST/ALT ratio, and a prolonged prothrombin time are the earliest indicators of cirrhosis and portal hypertension; however, the reliability of these markers requires further validation (3), and these indicators are unreliable in patients who have been treated with interferon (14). No serum markers that accurately predict the degree of hepatic fibrosis have been identified (3). Therefore, liver histology is frequently considered the gold standard for assessing hepatic fibrosis. However, liver biopsy is associated with sampling errors, interobserver variability, and potential complications (2), and the liver specimen obtained by the needle biopsy may not be representative of the overall degree of fibrosis. Thus, an urgent need exists for a simple and reliable noninvasive means of assessing disease severity in patients with chronic liver disease.

MRI, with the addition of MPG pulses, is presently considered to be the only means of performing in vivo measurements of molecular diffusion. Diffusion is the term used to describe the random thermal motion of molecules, called Brownian motion. Random movement in vivo includes not only pure water diffusion, but also other random microscopic motions, such as microcirculation or perfusion in the capillary network (4). Thus, the ADC, rather than the diffusion coefficient, is used to measure diffusion (5). This imaging technique has been used to detect cerebral ischemia at an early stage (15–17). With the advent of echoplanar imaging techniques, very high-speed imaging is now possible, and the use of echo-planar imaging with MPG pulses allows DW images to be obtained within a single breath hold, enabling the measurement of diffusion in abdominal organs.

Tissue ADC is thought to be composed of the mainly independent contributions of extracellular and intracellular tissue compartments (18). Concerning the contribution of the intracellular compartment, lower ADCs have been reported for components with a higher content of relatively impermeable impeding barriers, like cell and nuclear membranes, organelles, cytoskeleton, and matrix fibers (19). The early detection of ischemic brain is based on the observation of a decreased ADC caused by the reduced extracellular/intracellular volume ratio resulting from cytotoxic intracellular edema (16, 17). A reduction in intracellular proton movement arising from energy loss during ischemia has been considered as an alternate explanation. Concerning the contribution of the extracellular compartment, on the other hand, decreasing ADC levels with increasing radiation doses in rectal carcinoma were observed during combined, preoperative chemoradiation therapy (20). The limiting effect of fibers produced as a result of chemoradiation reduces free diffusion in the extracellular volume. Likewise, collagen fibers are not proton-rich, and their protons are tightly bound. Diffusion in regions of hepatic fibrosis, in which collagen fiber is the main component, may be similarly reduced, resulting in lower ADC values. In the present study, we hypothesized that the degree of hepatic fibrosis may be reflected by the ADC values calculated from DWI results, and that the increasing development of fibrous connective tissue in the liver would decrease the ADC values.

In the present study, the ADC values decreased as the stage of liver disease progressed from normal function to chronic hepatitis or cirrhosis, similar to the results obtained in most previous studies (6–11); however, the ADC values themselves varied, probably because of the different b values that were used. The ADC values tended to be higher when a pair of low b values was used because the decrease in signal intensities was affected not only by diffusion, but also by perfusion. ADC values using large b values, on the other hand, tended to be underestimated because of the greatly diminished image quality caused by the low signal intensity of the liver. In the present study, we chose two different, relatively low b values (0.01 and 128.01 seconds/mm2), and therefore the ADC values measured in this study would be affected not only by diffusion, but also greatly by perfusion.

In addition, we evaluated the relationship between the ADC values and the grade of fibrosis in the liver. We compared the ADC values and the fibrosis scores according to the HAI findings and found that the ADC values decreased as the fibrosis score increased, and the correlation was statistically significant. However, no relationships between the ADC values and the other HAI categories, like necroinflammation, were seen. ADC values might reflect the progression of fibrosis in the liver, but not the progression of inflammation or necrosis. On the other hand, Boulanger et al (12), whose report is the only one, that, to our knowledge, compared the ADC values with the histological score, reported that no correlation was found between the ADC values and the inflammation or fibrosis scores measured using the Ishak scale with five different b values, ranging between 50 and 250 seconds/mm2. The reason for the disagreement regarding the correlation between the ADC values and the fibrosis scores between our results and theirs is unknown. They selected 18 hepatitis C patients who were asymptomatic and were classified as stage A on the Child's clinical scale, while we selected patients of various clinical disease stages. Furthermore, we found that the ADC values decreased as the Child-Pugh scores increased, and significant difference in the ADC values among Child-Pugh stages was seen, similar to the observations of Aube et al (11) using b values of 200 and 400 seconds/mm2. As fibrotic changes, including narrowing sinusoids and decreased blood flow (restricting water mobility), progress, the ADC values can be expected to decrease. The results of the present study showed that the measurement of ADC values is potentially useful for the diagnosis and evaluation of fibrosis with regard to both the histological score and the clinical stage.

The present study has several limitations. First, MPG pulses were applied only along the z-axis. However, Taouli et al (10) showed that the liver has an isotropic diffusion pattern, probably because of its randomly organized structure, using diffusion gradients along three directions. This information may indicate that the use of multidirectional diffusion gradients is probably unnecessary for the design of hepatic diffusion studies. Second, the patient population, especially the histologically confirmed patient population (N = 31), was relatively small. Although inflammation may increase the number of free protons in the liver parenchyma and may increase the ADC, a relationship between the ADC values and the necroinflammation scores was not seen in the present study. Further studies involving a larger number of patients who had undergone both liver biopsies and DWI are needed to evaluate the correlation between ADC values and histological scores. Third, because two separate DWI series, not in one breathhold, were scanned with two different b factors, there could be slight misregistration between two DWIs. Although the ROIs were carefully positioned to be in the same region between two corresponding images, the possible imaging misregistration might have a slight effect on ADC values measured. If well-registered DWIs were acquired, a more quantitative analysis could be done, which may result in closer correlation among the ADC values and histological scores.

In summary, our results suggest that the measurement of ADC values using DWI may be useful for fibrosis staging in the liver. Noninvasive ADC measurements may obviate the need for a liver biopsy for fibrosis staging in some patients.

Acknowledgements

We thank Yujiro Tanaka, MD for helpful discussion and Yasushi An-naka, RT, Shin-ichi Ohtani, RT, and Ayuko Ito, RT for MRI data acquisition.

Ancillary

Advertisement