Journal of Gastroenterology

, Volume 46, Issue 3, pp 350–358

Real-time tissue elastography as a tool for the noninvasive assessment of liver stiffness in patients with chronic hepatitis C

Authors

  • Hiroyasu Morikawa
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Katsuhiko Fukuda
    • Department of GastroenterologyPL Hospital
  • Sawako Kobayashi
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Hideki Fujii
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Shuji Iwai
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Masaru Enomoto
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Akihiro Tamori
    • Department of Hepatology, Graduate School of MedicineOsaka City University
  • Hiroki Sakaguchi
    • Department of Hepatology, Graduate School of MedicineOsaka City University
    • Department of Hepatology, Graduate School of MedicineOsaka City University
Original Article—Liver, Pancreas, and Biliary Tract

DOI: 10.1007/s00535-010-0301-x

Cite this article as:
Morikawa, H., Fukuda, K., Kobayashi, S. et al. J Gastroenterol (2011) 46: 350. doi:10.1007/s00535-010-0301-x

Abstract

Background

Although histopathological examination by “invasive” liver biopsy remains the gold standard for evaluating disease progression in chronic liver disease, noninvasive tools have appeared and have led to great progress in diagnosing the stage of hepatic fibrosis. The aim of this study was to assess the value of real-time tissue elastography, using an instrument made in Japan, for the visible measurement of liver elasticity; in particular, comparing the results with those of transient elastography (Fibroscan).

Methods

Real-time tissue elastography (RTE), transient elastography (Fibroscan), liver biopsy, and routine laboratory analyses were performed in 101 patients with chronic hepatitis C. The values for tissue elasticity obtained using novel software (Elasto_ver 1.5.1) connected to RTE were transferred to four image features, Mean, Standard Deviation (SD), Area, and Complexity. Their association with the stage of fibrosis at biopsy and with liver stiffness (kPa) obtained by Fibroscan was analyzed.

Results

Colored images obtained by RTE were classified into diffuse soft, intermediate, and patchy hard patterns and the calculated elasticity differed significantly between patients according to and correlated with the stages of fibrosis (p < 0.0001). Mean, SD, Area, and Complexity showed significant differences between the stages of fibrosis (Tukey–Kramer test, p < 0.05). In discriminating patients with cirrhosis, the areas under the receiver operating characteristic curves (AUC) were 0.91 for Mean, 0.84 for SD, 0.91 for Area, 0.93 for Complexity, and 0.95 for Fibroscan.

Conclusions

RTE is a noninvasive instrument for the colored visualization of liver elasticity in patients with chronic liver disease.

Keywords

Liver fibrosisTransient elastographyUltrasoundLiver biopsy

Introduction

Hepatitis C virus (HCV) infects approximately 170 million individuals worldwide, according to a report from the World Health Organization [1]. Chronic liver damage attributable to HCV infection results in hepatic fibrosis, which is characterized by an unusual accumulation of extracellular matrix materials produced by fibroblast-like cells including stellate cells in the hepatic parenchyma. Hepatic fibrosis progresses towards cirrhosis, an end-stage liver injury, leading to hepatic failure, hepatocellular carcinoma, and finally death. Thus, precise evaluation of the stage of chronic hepatitis C with respect to fibrosis has become an important issue to prevent the occurrence of cirrhosis and to initiate appropriate therapeutic intervention such as viral eradication using pegylated interferon (PEG-IFN) plus ribavirin [2].

Although liver biopsy is acknowledged as the gold standard for staging disease progression, there are some limitations, including its invasiveness, risk of complications, sampling error, variability in histopathological interpretation, and the reluctance of patients to submit to repeated examinations [3]. Because of these disadvantages, there is a growing shift in clinical practice to utilize or develop ‘noninvasive’ methodologies to reflect the stage of liver fibrosis.

Several noninvasive approaches have appeared, such as serum fibrosis markers, transient elastography (Fibroscan®; Echosens SA, Paris, France), and real-time tissue elastography (RTE). Serum fibrosis markers include direct tests, such as hyaluronic acid and type IV collagen, and indirect approaches, which detect alterations in hepatic function but do not directly reflect hepatic extracellular matrix metabolism; these include platelet counts, coagulation studies, and hepatic transaminases, or their combinations in indices/scores, such as the aspartate aminotransferase-to-platelet ratio index (APRI) [4, 5].

Transient elastography, which has been developed for the measurement of liver stiffness, is considered to reflect more directly than other means the fibrotic evolution of chronic liver trauma [610]. In 2005, Castera et al. and Ziol et al. reported that liver stiffness measurements could be useful in assessing the presence of significant fibrosis (F2–4) and in suggesting the presence of cirrhosis in cohorts of patients with chronic hepatitis C; the areas under the receiver operating characteristic curves (AUCs) ranged from 0.79 to 0.83 for the prediction of F2–4 and were over 0.95 for the identification of cirrhosis [11, 12]. Friedrich-Rust et al. [13] assessed the overall performance of transient elastography for the diagnosis of liver fibrosis by a meta-analysis which included fifty articles; the mean AUCs for the diagnosis of significant fibrosis, severe fibrosis, and cirrhosis were 0.84, 0.89, and 0.94, respectively. The limitations of this method have also been discussed; intraobserver agreement is influenced by variables such as body mass index (BMI, particularly when ≥28), hepatic steatosis, and flares of transaminases [1114].

RTE is a method developed in Japan for the visual assessment of tissue elasticity integrated in a sonography machine, based on a Combined Autocorrelation Method that calculates the relative hardness of tissue rapidly from the degree of tissue distortion and which displays this information as a color image. The distortion of tissue is color-coded according to its magnitude and superimposed translucently on a conventional B-mode image. This simultaneous display enables us to evaluate the anatomical correspondence between tissue elasticity and B-mode images. The RTE image is constructed by the amount of displacement of the reflected ultrasound echoes under compression. Ultrasound elastography does not demonstrate physical elasticity directly, but shows the relative degree of tissue strain when subtle compression is applied. In hard tissue, the amount of displacement of the reflected ultrasound echoes is low, whereas in soft tissue, the amount of displacement is higher because soft tissue can be compressed more than hard tissue [15, 16]. This technology has already been proved to be diagnostically valuable in detecting space-occupying lesions in the breast, prostate, and pancreas [1720]. Friedrich-Rust et al. [21] attempted to determine the elasticity of liver tissue in 79 patients with chronic viral hepatitis. They developed an elasticity score from the color-coded bit-map image produced by the computer program Mathlab version 6 (Math Works, Natick, MA, US). However, the diagnostic accuracy of this semi-quantitative assessment for the prediction of significant fibrosis (METAVIR scoring system ≥F2), severe fibrosis (≥F3), and cirrhosis (F4) was not satisfactory; the AUCs were 0.75, 0.73, and 0.69, respectively [21].

We report here an updated RTE system as a tool for the noninvasive assessment of liver stiffness in patients with chronic hepatitis C. In this new system, all pixel data in the colored image were transformed into a histogram and a binary image for more accurate quantification, using an exclusive software program.

Methods

Patients

Ten healthy adult volunteers with no evident liver disease were recruited after giving their oral informed consent. One hundred and one patients with chronic hepatitis C, whose disease was defined by the presence of serum anti-hepatitis C virus (HCV) antibodies and serum HCV RNA, with serum levels of alanine aminotransferase above the upper limit of normal, were included in this study. Percutaneous liver biopsy or laparoscopy was performed on the patients within 1 week following Fibroscan and RTE analysis at the Department of Hepatology, Osaka City University Hospital, between September 2007 and September 2009. The study protocol accorded with the Helsinki Declaration and was approved by the ethics committee of our institution. Patients were enrolled and liver biopsy was performed after informed consent was obtained.

Liver histology and quantification of liver fibrosis

Liver biopsy was carried out using a 15-gauge Tru-Cut needle biopsy apparatus (Hakko, Tokyo, Japan). The median length of liver samples obtained at biopsy was 18 mm (range 10–25 mm). Tissue specimens obtained by liver biopsy were fixed immediately in 10% formalin solution and embedded in paraffin. After cutting, sections were stained with hematoxylin and eosin or Azan Mallory and the stage of fibrosis and grade of inflammatory activity in the liver were determined according to the METAVIR scoring system [22, 23]. All biopsy specimens were examined independently by two experienced pathologists who were blinded to the clinical data and the measurements of liver stiffness. Fibrosis was staged on a 0–4 scale as follows: F0, no fibrosis; F1, portal fibrosis without septa; F2, portal fibrosis and few septa; F3, numerous septa without cirrhosis; F4, cirrhosis. The chronic hepatitis activity was graded as follows: A0, none; A1, mild; A2, moderate; and A3, severe.

Real-time elastography

We used ultrasonography (Hitachi EUB-8500; Hitachi Medical, Chiba, Japan) and an EUP-L52 Linear probe (3–7 MHz; Hitachi Medical) for real-time tissue elastography (RTE). This system is commercially available currently for the diagnosis of mammary neoplasms [17].

Patients were examined in a supine position with the right arm elevated above the head, and were instructed to hold their breath. The examination was performed on the right lobe of the liver through the intercostals, because liver biopsy and Fibroscan were also performed at the same site. The RTE equipment displays two images simultaneously; the RTE image showing the region of interest (ROI) as a colored area and the conventional B-mode image (Fig. 1a). An area was chosen where the tissue was free of large vessels and near the biopsy point. The measurement was fixed to a rectangle with 30 mm length × 20 mm breadth placed 5 mm below the surface of the liver (Fig. 1a). The color in the ROI was graded from blue to red (Fig. 1b). We stored the RTE images as moving digital images for 20–40 s. Then ten static images, captured by the observer at random from the moving images, using AVI2JPG v6.10 converter software (Novo, Tokyo, Japan), were analyzed using the novel software Elasto_ver 1.5.1 (which was developed and donated by Hitachi Medical) on a personal computer. Numerical values of pixels were from 0 to 255 (256 stepwise grading) according to color mapping from blue (0) to red (255), and a histogram of the distribution was generated (Fig. 1c). The scale ranged from red for components with the greatest strain (i.e., softest components) to blue for those with no strain (i.e., hardest components). Green indicated average strain in the ROI, and therefore intact liver tissue displayed a diffuse homogeneous green pattern. An appearance of unevenness in the color pattern was considered to reflect a change in the liver stiffness. For quantification, all pixel data in the colored image were transformed into a histogram and binary image (Fig. 1c, d).
https://static-content.springer.com/image/art%3A10.1007%2Fs00535-010-0301-x/MediaObjects/535_2010_301_Fig1_HTML.gif
Fig. 1

Procedure of image analyses for real-time tissue elastography (RTE). a The region of interest (ROI) was fixed to a rectangle with about 30 mm length × about 20 mm breadth with a 400–600 mm2 area placed 5 mm below the surface of the liver. b The color-coded images from the ROI of RTE were analyzed by the software Elasto_ver 1.5.1. The colors ranged from blue to red, indicating the relative gradients from hardness to softness. c Mean and Standard Deviation were calculated by a histogram, which was generated by 256 stepwise grading derived from the color image obtained in b. d Area and Complexity were calculated from the binary image. Area was derived from the percentage of white regions (asterisks, i.e., hard area). Complexity was calculated by the following equation, periphery2/Area. Median value of the data was kept as representative of RTE parameters

Colored RTE images are usually classified into several patterns in the diagnosis of breast disease [17]. In this study, we evaluated liver fibrosis as three patterns: a diffuse soft pattern, an intermediate pattern, and a patchy hard pattern. The diffuse soft pattern was a relatively homogeneously spread, light-green colored image (Fig. 2a; the corresponding histology is shown in Fig. 2d). The intermediate pattern was typified by a partially mottled, dotted image with blue spots on a light green background (Fig. 2b; the corresponding histology is shown in Fig. 2e). The patchy hard pattern comprised mixed images with a patchwork effect of light green, red, and blue (Fig. 2c; the corresponding histology is shown in Fig. 2f).
https://static-content.springer.com/image/art%3A10.1007%2Fs00535-010-0301-x/MediaObjects/535_2010_301_Fig2_HTML.gif
Fig. 2

Representative colored images of real-time tissue elastography. a A case of F1 whose histology is shown in d. Relatively homogeneous image colored light green indicates a diffuse soft pattern in RTE. Liver stiffness measured by transient elastography (Fibroscan) was 4.1 kPa. b A case of F3 whose histology is shown in e. Partially mottled and dotted image with blue and red spots in the light green background indicates an intermediate pattern in RTE. Liver stiffness measured by Fibroscan was 9.6 kPa. c A case of F4 whose histology is shown in f. Mixed image with light green, blue, and red colors indicates patchy hard pattern. Liver stiffness measured by Fibroscan was 36.3 kPa. g Correlation of the averages of liver stiffness measured by Fibroscan with the three patterns of RTE images. The transition of these three patterns correlated positively with the liver stiffness (p < 0.01). df H&E staining. Yellow bars 2 mm

Transient elastography

Liver stiffness was also measured by transient elastography (Fibroscan®; Echosens SA, Paris, France). Briefly, patients were placed on the bed in the horizontally supine position. A probe was placed on the skin above the right intercostal space. The velocity of shear waves, which were generated temporarily and passed though the liver, was combined with Young’s modulus for the automated calculation of elasticity [9]. Ten successful acquisitions were performed on each patient. The results that obtained ten valid measurements with a success rate of at least 60% and an interquartile range under 30% were considered successful. A median of 10 valid measurements was regarded as the liver stiffness for a given subject, and expressed in kilopascals (kPa).

APRI

The APRI was calculated as follows: aspartate aminotransferase (/upper limit of normal) × 10/platelet count (104/mm3).

Statistical analysis

Box plots were used to study the distribution of the RTE values according to the stage of fibrosis. The trends were evaluated using the Jonckheere–Terpstra test. The Tukey–Kramer test was used to compare the data between the groups. The diagnostic performance of RTE parameters and transient elastography was assessed with receiver operating characteristic (ROC) curves. The ROC curve is a plot of the sensitivity versus 1-specificity for all possible cutoff values. The most commonly used index of accuracy is the area under the receiving operating characteristic curve (AUC), with values close to 1.0 indicating high diagnostic accuracy. Analyses were performed using JMP-8 software (SAS Institute, Cary, USA).

Results

Patients

The characteristics of our 101 patients with chronic HCV infection at the time of RTE and transient elastography are summarized in Table 1. In 91 of them, liver biopsy was performed successfully. Ten patients, who had no clinically overt sign of decompensated cirrhosis, were diagnosed with cirrhosis (F4) by the appearance of the liver surface at laparoscopy without liver biopsy. RTE was performed successfully in all patients, but five patients (F1, 2; F3, 1; F4, 2) failed to receive transient elastography measurements due to obesity and liver atrophy.
Table 1

Characteristics of the patients at the time of real-time tissue elastography examination

Characteristics

Patients (n = 101)

Sex: male/female

43/58

Age (years)

54 ± 13 (range 24–80)

BMI (kg/m2)

22.1 ± 3.1 (range 15.1–32.3)

Platelet count (×104/mm3)

16.4 ± 6.6

Total bilirubin (mg/dL)

0.9 ± 0.4

Prothrombin time (INR)

1.02 ± 0.1

ALT (IU/L)

58.2 ± 41.2

Fibrosis stage

 F0

6

 F1

48

 F2

15

 F3

16

 F4

16

Histological activity of 91 patients with liver biopsy

 A0

2

 A1

24

 A2

45

 A3

20

Values are means ± SEM

BMI body mass index, ALT alanine aminotransferase

Representative images of real-time tissue elastography

The results were described as the Mean, which indicates the mean of the histogram and ranges from 73.0 to 139.8 (median 111.9), and Standard Deviation (SD), which indicates the standard deviation of the histogram and ranges from 36.5 to 76.6 (median 60.8). In another analysis, the data were transformed into a binary image and the results were described as Area, which indicates the percentage of hard tissue and represents the hard tissue domain divided by the ROI and ranges from 5.0 to 54.7% (median 24.8%), and Complexity, which indicates the complex ratio of the shape of an extracted hard tissue domain in the ROI and is calculated as periphery2/area of the hard tissue domain and ranges from 15.9 to 40.21 (median 22.9) (Fig. 1d). The four image features were calculated automatically by Elasto_ver 1.5.1 (Hitachi Medical). Mean, SD, and Complexity were described in arbitrary units [a.u.].

The colored RTE images were classified into three patterns according to their visual appearance. The values for liver stiffness measurement by transient elastography (Fibroscan) were 6.9 ± 4.5, 10.9 ± 6.8, and 26.0 ± 15.8 kPa in the diffuse soft pattern group, intermediate pattern group, and patchy hard pattern group, respectively. Thus, these three patterns correlated significantly with the kPa values obtained by transient elastography (Jonckheere–Terpstra test, p < 0.0001) (Fig. 2g).

Relationship between real-time tissue elastography and histological parameters

Figure 3a–d shows box plots of the RTE values corresponding to fibrosis stage and includes the healthy volunteer (HV) group. The Mean decreased with increasing fibrosis score (Jonckheere–Terpstra test, p < 0.0001). SD, Area, and Complexity increased with increasing fibrosis score (Jonckheere–Terpstra test, p < 0.0001). The significant differences between each fibrosis stage were as follows: HV versus F3 and F4, F1 versus F3 and F4, and F2 versus F4 at every parameter; HV versus F2 at Mean, SD, and Area; F1 versus F2 at Mean; F2 versus F3 at Mean; F3 versus F4 at Complexity (Tukey–Kramer test, p < 0.05) No significant difference was found between the chronic hepatitis activity grades with same fibrosis stage at all parameters (Tukey–Kramer test).
https://static-content.springer.com/image/art%3A10.1007%2Fs00535-010-0301-x/MediaObjects/535_2010_301_Fig3_HTML.gif
Fig. 3

Parameter analyses measured by real-time tissue elastography (RTE) for each fibrosis stage. Box plots of each RTE value corresponding to fibrosis stages F1–4 and the healthy volunteer group (HV). The tops and bottoms of the boxes indicate the 1st and 3rd quartiles. The length of the box represents the interquartile range within which 50% of values are located. The lines through the middles of the boxes represent the medians. a Mean, b SD, c Area, and d Complexity. HV, n = 10. F1–4, n = 95. *p < 0.01, and **p < 0.05

Relationship between real-time tissue elastography and liver stiffness

Figure 4a–d, shows linear regression analysis of the values obtained by RTE compared to the liver stiffness values (kPa) obtained by transient elastography (Fibroscan). Simple regression analyses indicated that Mean, SD, Area, and Complexity were all significantly correlated with liver stiffness measured by Fibroscan (Mean: r = −0.585, p < 0.001; SD: r = 0.425, p < 0.001; Area: r = 0.590, p < 0.001; Complexity: r = 0.532, p < 0.001).
https://static-content.springer.com/image/art%3A10.1007%2Fs00535-010-0301-x/MediaObjects/535_2010_301_Fig4_HTML.gif
Fig. 4

Correlation between liver stiffness measured by transient elastography (Fibroscan) and the parameters of real-time tissue elastography. a Mean was negatively correlated with liver stiffness (kPa) (p < 0.01). Correlation coefficient was −0.585. b SD was significantly correlated with liver stiffness (kPa) (p < 0.01). Correlation coefficient was 0.425. c Area was significantly correlated with liver stiffness (kPa) (p < 0.01). Correlation coefficient was 0.590. d Complexity was significantly correlated with liver stiffness (kPa) (p < 0.01). Correlation coefficient was 0.532 (n = 96). a.u. arbitrary units

Relationship between real-time tissue elastography and platelet count, APRI, and other laboratory parameters

Simple regression analyses indicated that all Mean, SD, Area, and Complexity values were significantly correlated with the platelet count (Mean: r = 0.432, p < 0.001; SD: r = −0.332, p = 0.001; Area: r = −0.402, p < 0.001; Complexity: r = −0.393, p < 0.001). In addition, simple regression analyses indicated that Mean, SD, Area, and Complexity were all significantly correlated with APRI (Mean: r = −0.442, p < 0.001; SD: r = 0.373, p < 0.001; Area: r = 0.425, p < 0.001; Complexity: r = 0.418, p < 0.001). Furthermore, the correlation coefficient was significant for prothrombin time (Mean: r = −0.404, p < 0.001; SD: r = 0.343, p < 0.005; Area: r = 0.435, p < 0.001; Complexity: r = 0.440, p < 0.001), while no significant correlation was found for the four image features and total bilirubin, age, BMI, or alanine aminotransferase.

Diagnostic value of real-time tissue elastography and transient elastography

Figure 5 shows the ROC curves of RTE parameters for no significant fibrosis (F0–1) and cirrhosis (F4) in ninety-six patients who were also examined successfully by transient elastography. The AUCs for the stage of no significant fibrosis (F0–1) were 0.89, 0.81, 0.87, and 0.81 for Mean, SD, Area, and Complexity, respectively. The AUCs for severe fibrosis (≥F3) were 0.93, 0.84, 0.91, and 0.86 for Mean, SD, Area, and Complexity, respectively. The AUCs for cirrhosis (F4) were 0.91, 0.84, 0.91, and 0.93 for Mean, SD, Area, and Complexity, respectively. In transient elastography (Fibroscan), the AUCs were 0.92, 0.95, and 0.95 for stages F0–1, ≥F3s and F4, respectively. The corresponding sensitivities, specificities, and positive and negative predictive values are detailed in Table 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs00535-010-0301-x/MediaObjects/535_2010_301_Fig5_HTML.gif
Fig. 5

Receiver operating characteristic curves of each parameter obtained by RTE. a No significant fibrosis (F0–1). b Cirrhosis (F4). a The areas under the receiver operating characteristic curves (AUC) for no significant fibrosis (F0–1) were 0.89, 0.81, 0.87, 0.81, and 0.92 for Mean (red), SD (blue), Area (yellow), Complexity (pink), and transient elastography (Fibroscan, black), respectively. b The AUCs for cirrhosis were 0.91, 0.84, 0.91, 0.93, and 0.95 for Mean, SD, Area, and Complexity, and transient elastography, respectively (n = 96)

Table 2

Cutoff values of real-time tissue elastography (image features) and transient elastography for the diagnosis of fibrosis stages (F)

 

F = 0–1

F ≥ 3

F = 4

Mean (AUC)

0.89

0.93

0.91

 Cutoff (a.u.)

110.1

106.9

101.5

 Sensitivity (%)

84.1

82.8

85.7

 Specificity (%)

82.7

85.1

82.9

 Positive predictive value (%)

80.4

70.6

46.2

 Negative predictive value (%)

86.0

91.9

97.1

SD (AUC)

0.81

0.84

0.84

 Cutoff (a.u.)

61.2

63.0

65.7

 Sensitivity (%)

70.5

75.9

78.6

 Specificity (%)

73.1

77.6

79.3

 Positive predictive value (%)

68.9

59.5

39.3

 Negative predictive value (%)

74.5

88.1

95.6

Area (AUC)

0.87

0.91

0.91

 Cutoff (%)

25.8

29.5

33.7

 Sensitivity (%)

81.8

79.3

85.7

 Specificity (%)

80.8

80.6

86.6

 Positive predictive value (%)

78.3

63.9

52.2

 Negative predictive value (%)

79.6

90.0

97.3

Complexity (AUC)

0.81

0.86

0.93

 Cutoff (a.u.)

23.2

24.9

27.8

 Sensitivity (%)

77.3

79.3

85.7

 Specificity (%)

75.0

80.6

87.8

 Positive predictive value (%)

72.3

63.9

54.5

 Negative predictive value (%)

79.6

90.0

97.3

Transient elastography (AUC)

0.92

0.95

0.95

 Cutoff (kPa)

10.1

13.3

16.3

 Sensitivity (%)

88.6

89.7

85.7

 Specificity (%)

86.5

86.6

85.4

 Positive predictive value (%)

84.8

74.3

50.0

 Negative predictive value (%)

90.0

95.1

97.2

AUC the area under the receiver operating characteristic curves, a.u. arbitrary units

Discussion

Recently, various techniques based on ultrasound or magnetic resonance imaging have been developed to quantify liver stiffness, and thereby liver fibrosis, noninvasively. Among them, transient elastography (Fibroscan) has been used most frequently worldwide and has become established in clinical practice to detect advanced fibrosis without liver biopsy, although several limitations and disadvantages of the modality have been discussed [8, 24]. Another novel imaging modality is magnetic resonance elastography (MRE). The technique typically is added to a conventional MR examination of the upper abdomen [25]. A pneumatic or electromechanical driver is placed in contact with the abdominal wall and is used to generate propagating mechanical waves in the liver at frequencies between 40 and 120 Hz. Although MRE was shown to be superior to APRI and transient elastography for determining the stage of fibrosis in patients with various underlying liver diseases [26], MRE cannot be performed on an iron-overloaded liver because of noise. In addition, MRE takes a longer time and costs more than the ultrasound-based elastographies [2].

We paid attention in our analysis to the pattern change of RTE color images according to the progression of fibrosis. Normal or minimally fibrotic liver exhibited a homogeneous RTE image that was colored light green (Fig. 2a). According to the progression of liver fibrosis, the homogeneous pattern transited to a patchy pattern consisting of a blue-colored area (Fig. 2c), which may suggest a collapse of homogeneity. In the present study, for semi-quantification of the RTE image, we used a histogram and binary image produced using an exclusive software program that was developed by Hitachi Medical. This is the first report demonstrating the utility of Mean, SD, Area, and Complexity as RTE parameters. We speculate that Mean and Area may directly represent liver elasticity, while SD and Complexity may imply the collapse of the uniform architecture of the liver parenchyma concomitant with progressing hepatic fibrosis (Figs. 3, 5).

After the report by Friedrich-Rust et al. [21] other investigators criticized the intraobserver variability and the lack of interobserver agreement in hepatic RTE [2729]. In general practice, an operator presses lightly on the surface of the liver through the skin with a transducer when the elastogram is generated. Thus, the pressure generated by the operator’s compression is assumed to influence both the image of elasticity and the resulting elasticity score. To avoid this source of error, we used here the latest and most sensitive probe that was produced by Hitachi Medical and did not require extra external stress. Accordingly, we were able to improve the acquisition of the color image representing the distortion of liver tissue under the heartbeat or abdominal aorta. We also adopted ten individual frames for semi-quantitative analysis. On the other hand, Saftoiu et al. and Gulizia et al. have proposed that the ROI should include the surrounding tissues, such as adipose tissue, diaphragm, and intercostal muscles, in order to clearly compare and distinguish the strain between the liver and these organs [27, 30]. However, we placed the ROI inside the liver at 5 mm below the surface of the liver, because the new probe used in this study is sensitive and the deep attenuation of the ultrasound image could be disregarded. We avoided including the liver surface inside the ROI because the liver surface is hard and therefore is assessed as a harder area, influencing the histogram analysis. Recently, Tatsumi et al. [31] also reported the results of RTE using the ROI inside the liver in a similar fashion to our study.

We note that all four image features of RTE, comprising Mean, SD, Area, and Complexity, were significantly correlated with the kPa value obtained by Fibroscan [8, 1113]. In particular, Mean and Area had a high correlation. In addition, the AUC values were similar between RTE and Fibroscan. Mean and Area were highly accurate for the diagnosis of significant fibrosis (i.e., >F0–1) and for the diagnosis of cirrhosis (F4). Although the performance of Fibroscan has been demonstrated in many studies to have high accuracy [4, 8], the machine is used solely for elasticity measurements. In contrast, with the new equipment used in the present study, RTE can be used simultaneously with conventional B-mode ultrasonography. Moreover, as reported by Obara et al. [32], liver stiffness measurement by Fibroscan was unsuccessful in 5.3% of Japanese cases of chronic liver disease, similar to our experience in the present study (5.0%). Thus, RTE is considered to be superior to Fibroscan at points where measurements by Fibroscan are difficult to perform in obese patients and are impossible to perform in patients with ascites [32, 33], and where RTE images are unaffected by steatosis, as suggested previously by Friedrich-Rust et al. [29]. Furthermore, because Fibroscan measures liver stiffness between 25 and 65 mm below the surface of the skin [10], knowledge of the relative thickness of the liver is necessary for the measurements.

The diagnostic performance of RTE is similar to that of other noninvasive laboratory tests, such as the FibroTest (BioPredictive, Paris, France), APRI, and the Forns score reported in the literature [4, 8]. The FibroTest is based on five serological parameters; bilirubin, gamma-glutamyl transpeptidase (GGT), apolipoprotein A1, α2-macroglobulin, and haptoglobin. While the diagnostic accuracy was high (AUC 0.87) for significant fibrosis (METAVIR, ≥F2), the FibroTest costs more than APRI and the Forns score, and needs two uncommon parameters [34]. In APRI, using the cutoffs proposed by Wai et al. [5], approximately 50% of patients could be correctly classified as having cirrhosis without a liver biopsy5. With the Forns’ index, the AUC for the prediction of significant fibrosis (Scheuer classification, ≥F2) was 0.86 in the test set and 0.81 in the validation set [35]. It is, however, known that the determination of the severity of liver fibrosis by serum markers is confounded by acute inflammation, hemolysis, cholestasis, and renal failure [4, 5, 34, 35]. Castera et al. [11] compared the performance of transient elastography with that of the FibroTest, APRI, and liver biopsy for the assessment of liver fibrosis in a large number of patients with hepatitis C. Interestingly, they reported that the best performance was achieved by a combination of transient elastography with the FibroTest [11]. Friedrich-Rust et al. reported that the best diagnostic accuracy was obtained by combining the variables used for the calculation of the RTE elasticity score with the platelet count and GGT [18]. Thus, RTE, in combination with serological parameters, can further improve the accuracy of differentiating fibrosis stages.

One of the major limitations of the present study was that the number of patients with F1 was higher than the number of those with the other stages, because most of our patients received liver biopsy prior to interferon treatment. However, our study compared the performance of RTE with that of transient elastography in the same patients. Although the AUC for RTE in this study was higher than that in the studies by Friedrich-Rust et al. [21, 29], the AUC for transient elastography was approximately equivalent to that in one of these studies [29].

In the METAVIR and Desmet’s histological scoring systems, cirrhosis is classified as a single category (i.e., F4) [22, 23]. However, the degree of fibrosis; for example, the content of collagen and extracellular matrix materials that may be closely associated with the function of hepatocytes and portal hypertension, may vary among patients with cirrhosis. Foucher et al. [36] reported that the kPa measured by transient elastography in cirrhotic patients correlated well with clinical parameters indicating the severity of cirrhosis; 27.5 kPa was the cutoff value for the presence of esophageal varices stage 2 or 3, 37.5 kPa for liver function Child B or C, 49.1 kPa for a past history of ascites, and 62.7 kPa for esophageal variceal bleeding. Thus, because cirrhosis can be staged in greater detail with clinical relevance based on liver stiffness with RTE, RTE may be useful for this staging of cirrhosis and for detecting and assessing the risk of cirrhotic complications [36].

In summary, we have shown a convenient and noninvasive procedure, RTE, for the visual assessment of liver stiffness. The performance of RTE compares favorably with that of transient elastography (Fibroscan) for detecting the presence of significant liver fibrosis in patients with chronic hepatitis C. We suggest that RTE could be used as a routine imaging method to evaluate the degree of liver fibrosis in patients with liver disease. Future studies of larger patient cohorts will be necessary for the validation of the four RTE parameters, and the combination of these parameters will enable improvement of accuracy in assessing hepatic fibrosis.

Acknowledgments

We thank Ms. Akiko Tonomura and Mr. Junji Warabino, Hitachi Medical Co., for the technical support for RTE. Hiroyasu Morikawa was supported by a research grant from Osaka City University (2009). Norifumi Kawada was supported by Research on Hepatitis, Health and Labour Science Research Grants from the Ministry of Health, Labour and Welfare of Japan (2008–2009) and by a Thrust Area Research Grant from Osaka City University (2008–2009).

Copyright information

© Springer 2010