A pilot approach for quantitative assessment of liver fibrosis using ultrasound: preliminary results in 79 cases

Journal of Hepatology 44 (2006) 68–75 www.elsevier.com/locate/jhep

A pilot approach for quantitative assessment of liver fibrosis using ultrasound: preliminary results in 79 cases Hiroyuki Yamada1,*, Masaaki Ebara1, Tadashi Yamaguchi5, Shinichirou Okabe1, Hiroyuki Fukuda1, Masaharu Yoshikawa1, Takashi Kishimoto2, Hisahiro Matsubara3, Hiroyuki Hachiya4, Hiroshi Ishikura2, Hiromitsu Saisho1 1

Medicine and Clinical Oncology, Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuou-ku, Chiba 260-8670, Japan 2 Molecular Pathology, Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuou-ku, Chiba 260-8670, Japan 3 Academic Surgery, Graduate School of Medicine, Chiba University, 1-8-1 Inohana, Chuou-ku, Chiba 260-8670, Japan 4 Research Center for Frontier Medical Engineering, Faculty of Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan 5 Department of Information and Image Sciences, Faculty of Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan

Background/Aims: Ultrasound is noninvasive and useful to evaluate liver disease despite its operator dependency. This pilot study was conducted to quantitatively assess liver fibrosis using ultrasound. Methods: Fibrosis extraction ratios (FER) (fiber volume/total volume) of ultrasound and histological images of 8 autopsy specimens were compared. We also compared FER of ultrasound images from clinical patients (nZ79) with histological fibrosis stages. Results: In the autopsy study, FER correlation coefficient between histological images and ultrasound images was 0.992. Regarding clinical patients, there was sufficient evidence to indicate differences in the distributions of FER for each fibrosis stage (Kruskal–Wallis test P!0.0001). With FER cut-off to distinguish RF2 from F0 and F1 defined as mean plus standard deviation of F0 and F1, sensitivity, specificity, positive predictive value, negative predictive value, and likelihood ratio were 62, 75, 78, 57%, and 2.47, respectively. Regarding HCV cohort (nZ44), they were 55, 87, 89, 50%, and 4.14, respectively. Areas under receiver operating characteristic curves were 0.78, 0.79, 0.83 and 0.83 for RF1, RF2, RF3 and ZF4, respectively. Regarding HCV cohort, they were 0.74, 0.71, 0.79 for RF2, R3 and Z4, respectively. Conclusions: The FER method has great potential for diagnosing liver fibrosis using ultrasound. q 2005 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Keywords: Liver fibrosis; Quantitative; Ultrasound; Pilot

1. Introduction At present, percutaneous liver biopsy is the gold standard in assessing liver fibrosis, but there could be sampling error [1,2], and specimens might not represent the state of the whole liver accurately because only about 0.002% of the organ is sampled. Regarding the assessment of specimens,

Received 26 November 2004; received in revised form 17 August 2005; accepted 18 August 2005; available online 15 September 2005 * Corresponding author. Tel.: C81 43 226 2083; fax: C81 43 226 2088. E-mail address: [email protected] (H. Yamada).

there are various histological evaluation systems [3,4], and the results may depend on the histologist’s experience. Several investigators have reported serological approaches for the quantitative diagnosis of liver fibrosis [5–8]. Some investigators have also used imaging procedures such as CT imaging, MR imaging and ultrasound in fibrosis assessment [9,10]. Among them, ultrasound is especially noninvasive and easily performed, and there are many reports about the quantitative estimation of diffuse liver disease using ultrasound [11–21]. We have also studied the ultrasound patterns of liver fibrosis using a neural network [22]. However, reliable methods for the quantitative analysis of liver fibrosis are at present not available.

0168-8278/$30.00 q 2005 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jhep.2005.08.009

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75

The objectives of the present study were to compare ultrasound images with histological images of autopsy liver specimens, and to compare ultrasound image analysis with the results of histological evaluation in patients who had liver biopsies or partial liver resection, with regard to liver fibrosis using a pilot ultrasound image analysis system [23,24].

2. Patients and methods 2.1. Autopsy study 2.1.1. Autopsy materials We studied eight established cirrhotic liver specimens (etiology is shown in Table 1) obtained at autopsy during 1998–2001 at the 1st Department of Internal Medicine, Chiba University Hospital. Of the 8 cases, 6 had hepatocellular carcinoma (HCC), and we used specimens (about 50 mm!50 mm!20 mm) not including macroscopically identified HCC. All related families gave their written informed consent.

2.1.2. Ultrasound data acquisition and extraction of fiber information We placed the liver specimens in a test tank filled with degassed water and scanned in the direction of 20-mm thickness using ultrasonic diagnostic equipment (SSA 770A, Toshiba Medical Systems, Tokyo, Japan). We attached a probe to a precision motorized device that moves along a rail at a speed of 0.7 mm/s, and then placed the machine with the rail over the tank (the probe was in the water). The device with the probe, moving along the rail recorded 200 consecutive ultrasound images from each 20-mm-thick liver specimen. Transforming and receiving frequencies were 2.0 and 4.0 MHz, respectively. In previous studies, we analyzed the relationship of changes in echo characteristics and tissue [24], and we proposed a pilot technique for extracting information concerning diseased tissue in cirrhotic liver from the echo signals [23]. The signal of the speckle part (from normal liver tissue) was statistically approximated to Rayleigh distribution (Fig. 1). On the other hand, many of the echo signals from diseased tissue are independent from the speckle pattern, and they do not obey Rayleigh distribution (Fig. 2). In this report, they are termed Rayleigh element and non-Rayleigh element, respectively. Non-Rayleigh elements representing fiber structure are extracted by the following procedure. Rayleigh distribution is given by pðxÞ Z

2x Kðx2 =s2 Þ e s2

(1)

2

where s is the variance of the echo amplitude. If the echo signal that obeys Rayleigh distribution passes through a logarithmic amplifier, log-compressed output y is given by y Z k lnðlxÞ

(2)

where k and l are constants. The variance of y is calculated as the result of Table 1 Fibrosis extraction ratios (FER) of autopsy specimens from ultrasound images and histological images Etiology of cirrhosis

Ultrasound

Histology

Hepatitis B virus* Hepatitis C virus (HCV) HCV HCV HCV Alcoholic liver disease (ALD) ALD Secondary biliary cirrhosis*

0.1582 0.1029 0.1203 0.1451 0.1152 0.1578 0.1513 0.0941

0.2691 0.2747 0.2478 0.2982 0.1921 0.3176 0.3185 0.2174

All with hepatocellular carcinoma except *.

69

subtraction of the average of the signal y from y. v Z yKhyi

(3)

After the subtraction processing, the variance of output v will become a constant. Thus, the echo envelope of a normal liver is suppressed under the constant value. If the signals that reverse log-compressed are defined as z, z is given by z Z m env ;

(4)

where m and n are constants. In the result of signal processing from Eqs. (2)–(4), the speckle pattern that obeys Rayleigh distribution will be controlled under low amplitude. On the other hand, non-Rayleigh signals will emphasize high amplitude [23]. When Rayleigh and non-Rayleigh elements are both included in the input signal, our extraction technique determines the information of zOT as the non-Rayleigh signal. T is the quantitative threshold defined by Eq. (5). pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (5) T Z m KlnðPn Þeg : where g is the Euler number, and Pn is the probability of the echo information from normal tissue being recognized as non-Rayleigh element. This is called the false extract rate. If PnZ10K2, it means that the result of the extractions might include the 10% error that the Rayleigh element is extracted as non-Rayleigh element. In this report, Pn is fixed at 10K2 for all data. Fig. 3b–d shows the results of applying extraction processing to an original echo image (Fig. 3a). Fig. 3d might represent fiber structure.

2.1.3. Histological data acquisition and extraction of fiber information After ultrasound scanning, 20 histological preparations of 4-mmm thickness were obtained 1 mm apart from each 20-mm-thick liver specimen, and stained with Masson’s trichrome for fibrosis evaluation (Fig. 4a). Images were taken from the stained histological preparations by high-resolution scanner, and saved to a computer as tagged image file format (TIFF) images. Collagen fiber was stained with aniline blue in the known Masson’s trichrome stain, the information on fiber structures was simply separated by calculating the blue component minus the red component (Fig. 4b), and we confirmed the obtained results compared with the original histological preparations visually. The fibrosis stages of all specimens were F4 (cirrhosis) according to the New European Classification [25].

2.1.4. Reconstruction of the 3D structure of fiber tissue The extraction processing result of the 2D echo image was arranged on a scanning time-axis, and rendering was performed in six directions in each pixel of each frame. Consequently, the point that had been independent in the 3D direction (such as noise) would not connect to the other points. On the other hand, structural information (such as fiber tissue) would be combined. By this step, it became possible to view the tissue structure from arbitrary angles. Then, we constructed the fiber structure from the extracted results of histological images by the same method. 3D fiber structures were observed, and finally the ratios of fiber volume: extraction fibrosis ratio (FER) were calculated (fiber volume/total volume) in each 3D image for a quantitative assessment of fibrosis. Four regions of interest (10 mm! 10 mm!10 mm each) used for calculating FER were set on the 3D images not containing any large vessels, and the means of the four values were used for FER of each case. We utilized a generally used computer (Endeavor Pro 2000, Epson, Nagano, Japan) and software (MATLAB R14, The Math Works, Natick, USA) to extract the fiber components and calculate FER.

2.1.5. Statistical analysis Simple linear regression and scatter plot analysis were used to assess the relationship between FER of ultrasound images and those of histological images.

2.2. Patient study 2.2.1. Patients We studied 74 patients who underwent percutaneous liver biopsy by 18gauge needle with 20-mm specimen notch, and 5 patients who had partial liver resection because of malignancy (4 hepatocellular carcinoma,

70

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75

Fig. 1. Sample of normal liver: ultrasound image (a) and probability density function (PDF) of echo amplitude (b). Speckle pattern appears over the whole liver and PDF of echo envelope obeys Rayleigh distribution. 1 hepatic metastasis) between September 2003 and August 2004 at Chiba University Hospital. Only liver samples presenting at least 10 portal tracts were considered suitable for evaluation. They consisted of 41 males and 38 females, mean age 51G11 years (range, 19–70 years) and mean body mass index 23.9G4.7 Kg/m2 (range, 11.1–42.4 Kg/m2). The etiology was hepatitis C virus in 44, hepatitis B in 20, autoimmune hepatitis in 4, alcoholic liver disease in 3, primary biliary cirrhosis in 3, and other causes in 5 patients. All patients gave informed consent to all clinical investigations, performed in accordance with the principles of the Declaration of Helsinki.

2.2.2. Ultrasound data acquisition and extraction of fiber information We used the same ultrasonic diagnostic equipment and frequencies as for the autopsy study. We applied the transducer lengthways to the epigastric lesion of the patient’s body surface, moving it in a linear fashion along the patient’s skin manually about 3 cm for 100 consecutive ultrasound images. The patients held their breath during the scanning, usually only for 15 s. FER was calculated in each case by the same method as for the autopsy specimens for quantitative assessment of fibrosis. All ultrasound scanning was performed on the day prior to liver biopsy or a few days before partial liver resection. With regard to inter- and intra-observer variability of scanning, we used 10 patients. In the inter-observer study, the first investigator (H.Y.) performed scanning, followed by scanning by a second investigator (M.E.). In the intra-observer study, the investigator

(H.Y.) performed scanning for the first time, and then after a short break, for a second time. The ultrasound data analysis would show no variability because the procedure was done by computer. FER obtained in the interand intra-observer studies showed very low variability and good reproducibility. Thus, all subsequent scanning was done by one gastroenterologist (H.Y.).

2.2.3. Histological assessment Biopsied and resected specimens were evaluated with regard to inflammation activity and fibrosis in a blind fashion by two independent liver pathology specialists (T.K. and H.I.) based on the New European Classification. Steatosis was estimated according to the percentage of hepatocytes with fat droplets. If the percentage was greater than 30, we regarded it as steatosis. When their diagnoses were in disagreement, they re-examined the samples and reached consensus after discussion. For this study, the stages of fibrosis were scored as F0, no fibrosis; F1, mild fibrosis; F2, moderate fibrosis; F3, severe fibrosis; F4, cirrhosis.

2.2.4. Statistical analysis Box plots and Kruskal–Wallis test were used to study the FER distribution according to histological fibrosis stages. In addition, the Mann–Whitney U-test was used to delineate differences between respective fibrosis stages. Finally, receiver operating characteristic (ROC) analysis of FER was performed. The area under each ROC curve was estimated using the trapezoidal rule.

Fig. 2. Sample of cirrhotic liver: ultrasound image (a) and probability density function (PDF) of echo amplitude (b). Speckle pattern changes to typical ultrasound patterns appeared as high or low echo spots, and PDF of echo envelope does not obey the Rayleigh distribution.

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75

71

Fig. 3. Results of extraction technique of an ultrasound image. (a) Original ultrasound image; (b) Result of Eq. (2); (c) Result of Eq. (4); (d) Result of Eq. (5) extracted fiber image.

Fig. 4. Result of extraction technique of a histological image. (a) Original specimen stained with Masson’s trichrome stain. (b) Fiber image: result of extracting blue component from original specimen.

3. Results 3.1. Correlation between histological and ultrasound images of autopsy specimens The reconstructed results of an ultrasound image and a histological image are shown in Fig. 5: the structures were clearly visible to the surface and the interior of the liver.

The internal structure was net-like, fiber thickness was about 1–3 mm, and the size of the portion inside the net-like structure was 2–5 mm. From these features, it was considered that each provided the information of fiber and nodule. In order to observe the detailed structure, a characteristic part of tissue sized 20 mm!20 mm!20 mm was extracted from each 3D image, and they were compared in detail. As shown in Fig. 6, fiber structure, nodule size, and

Fig. 5. 3D fiber structures constructed from ultrasound fiber images (a) and histological fiber images (b). The two fiber structures are similar.

72

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75

Fig. 6. Detailed 3D fiber structures extracted from Fig. 5. The constructed fiber structure from histological fiber images (b) looks thicker than that from ultrasound fiber images (a).

position were almost equivalent. The fact that the fiber structures crossed intricately was also a similarity. FER of ultrasound images and histological images are shown in Table 1. Scatter plot and linear regression line are shown in Fig. 7. There was significant correlation between FER of histological images and those of ultrasound images (correlation coefficient 0.992). 3.2. Relationship between FER and histological fibrosis stage in patient study Fiber images were extracted from ultrasound images (Fig. 8). Histological fibrosis stage distribution among the patients and FER according to fibrosis stages are shown in Fig. 9. There was sufficient evidence to indicate differences in the distributions of FER for each stage (P!0.0001). We could not delineate significant differences between F0 and F1 (PZ0.162), and between F3 and F4 (PZ0.389), although between all other stages we could find significant differences (P!0.05). In contrast with this result, activity of

inflammation and steatosis showed no statistically significant differences in each stage of fibrosis. With regard to the patients with HCV infection, dividing them into those without significant fibrosis (F0 and F1) and those with advanced fibrosis for whom antiviral treatment should be performed (RF2) is important. Table 2 shows the results of sensitivity, specificity, positive predictive value, negative predictive value, and likelihood ratio, with the cutoff value defined as mean plus standard deviation of the patients without significant fibrosis among all patients and the HCV cohort. The ROC curves of FER were plotted for the histological fibrosis stages greater than or equal to F1, F2, F3 and for liver fibrosis stage at F4 (Fig. 10). The areas under the ROC curves, which estimate the performance diagnostic of FER, were 0.78, 0.79, 0.83 and 0.83 for liver fibrosis stages greater than or equal to F1, F2, F3 and F4, respectively.

FER of Histological images

.34 .32 .3 .28 .26 .24 .22 .2 .18 .09

.1

.11

.12

.13

.14

.15

.16

FER of Ultrasound images Fig. 7. Relationship between the fibrosis extraction ratios (FER) of histological images and those of ultrasound images. Scatter plot and linear regression line of 8 autopsy specimens are shown. Correlation coefficient: 0.992; slope of the line (regression coefficient): 2.019.

Fig. 8. 3D fiber structure constructed from ultrasound images obtained by transabdominal wall scanning. The shape of liver and fiber structures inside are well represented.

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75 .08 .07

FER

.06 .05 .04 .03 .02 Kruskal-Wallis test p<0.0001

.01 0

1

2

3

4

(6)

(26)

(24)

(12)

(11)

Fibrosis stage Fig. 9. Fibrosis extraction ratios (FER) for each fibrosis stage. The top and bottom of the boxes are 75th and 25th percentiles. The line through the middle of the box represents the median (50th percentile). The upper and lower adjacent lines are 90th and 10th percentiles. All dots above and below these lines represent observations O90th and !10th percentile, respectively. Figures in parentheses under each fibrosis stage represent the number of patients.

With regard to the HCV cohort, they were 0.74, 0.71, and 0.79 for liver fibrosis stages greater than or equal to F2, F3 and F4, respectively (the HCV cohort contained no F0 patients).

4. Discussion As is well known, the volume of fiber in the liver is highly correlated with the degree of liver fibrosis as assessed by biopsy. Some investigators used histological 2D images obtained from liver biopsy to perform digital quantification of the fibrous area, reporting good correlations between the results and fibrosis scoring by pathologists [26–28]. Such quantitative studies elicit objective results that would not depend on investigators. However, this may entail two problems, namely, that because the samples used in these studies were obtained through biopsy, the image size was very small and they might not have been representative of the entire liver, and that these methods required liver biopsy, to a certain extent an invasive technique. Our in vivo study has three advantages in comparison with the previously published studies. First, we used considerably larger-size 3D volumetric data compared with needle biopsy. So the sample data might represent

73

the entire liver. Second, the results were expressed numerically, meaning that they are convincingly objective. Third, the method used in the patient study requires no invasive technique and acquisition of the ultrasound data can be done in very short order. On the other hand, our in vivo method has a weak point in that the geometric measurements of the 3D data were not exactly accurate because the transducer is moved untracked and freehand. So our method can certainly stand further improvement. Some investigators used mechanical attachments that move at predetermined spatial or angular intervals with the transducer to make 3D images [29]. They further reported that, although these attachments were more cumbersome for the users, they might also give more geometric accuracy to 3D data. With regard to the autopsy specimens, FER of histological images were larger than those of ultrasound images in all specimens (Table 1). We used severe cirrhosis autopsy specimens with thick fiber structures. Ultrasound has strong signals, where the structure borders others (in this study, fiber bordered nodules). Our extraction method might not extract signals if the regions of interest have homogeneous tissue (in this study, especially inside thick fiber of the autopsy specimens), and the fibrosis images made from ultrasound images might represent the surface of thick fiber structures while those made from histological images showed the whole fiber structures (Fig. 6). Further, histological images contained more minute information than ultrasound images, so extracted fiber images from histological images could reveal more fiber structures than those of ultrasound images. This may be the reason for the disparity between FER of histological images and those of ultrasound images. In the patient study, FER from ultrasound images were smaller than those from autopsy specimens. There might be two reasons for this disparity. First, all biopsy and resected materials were not severe cirrhosis (Child-Pugh classification A, except for 3 B), while all autopsy materials had been established as severe cirrhosis (Child-Pugh classification C), and the volume of fibers was greater in the autopsy materials. Second, the ultrasound images obtained by transabdominal wall scanning might have encountered some interference by the abdominal wall, while the water tank study should have produced little interference. In this pilot study, at first we compared ultrasound fibrous images with histological fibrous images of autopsy liver specimens. Although histological FER were larger than

Table 2 Performance characteristics for distinguishing SF2 from F0 and F1

All patients (nZ79) HCV (nZ44)

Cut-off

Sensitivity (%)

Specificity (%)

PPV

NPV

Likelihood ratio

0.038 0.04

62 55

75 87

78 89

57 50

2.47 4.14

The cut-off value of the fibrosis extraction ratio for distinguishing FS2 from F0 and F1 defined as mean plus standard deviation of F0 and F1. PPV, positive predictive value; NPV, negative predictive value.

74

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75

cirrhosis is considerable. Our advantage is that we used larger-size 3D volumetric data to analyze FER and the results might represent the entire liver, although we need further improvement and a large cohort. Nonetheless, although the studied population size was too limited to define absolute thresholds, our preliminary results show that the FER method can divide patients into different fibrosis stages, and our method has a great potential for diagnosing liver fibrosis using only ultrasound.

Sensitivity

1

0.5

F1 F2 F3 F4 0

0

0.5

(0.78) (0.79) (0.83) (0.83)

References

1

1 - Specificity Fig. 10. The receiver operating characteristic (ROC) curves of fibrosis extraction ratios (FER) from ultrasound for the prognostic of fibrosis stage greater than or equal to F1, F2, F3 and F4. The areas under each ROC curve are indicated in parentheses.

those of ultrasound in all cases, they were well correlated. Following this, we compared FER of in vivo ultrasound images by transabdominal wall scanning with histological fibrosis stages of liver specimens. There was a significant difference of FER among the respective fibrosis stages. In a clinical setting, the results are expressed numerically, and objectively, as the extraction ratio, meaning that the viewer’s experience or technique may be largely irrelevant. This method requires no invasive technique, acquisition of the ultrasound data can be done in very short order after usual abdominal ultrasound examination, and thus patients will suffer minimal inconvenience and discomfort. For example, patients undergoing interferon therapy could be examined frequently to assess fibrosis. Another benefit, of course, is that this method requires no additional special device, meaning that it can be a very low-cost examination. For the FER method, the clinician at bedside would change the ultrasound mode to the FER setting and save pictures in the usual way. As calculation software is not yet installed in the ultrasonic diagnostic equipment, data is moved to a computer and FER are calculated. On the other hand, a disadvantage acknowledged in the present study is that it was difficult to define the absolute threshold for dividing patients into their respective fibrosis stages. Recently, some papers have used transient elastography (FibroScanw) to assess liver fibrosis [30,31]. The results are expressed numerically, and it also requires no invasive technique. Their AUCs were almost the same as ours for detecting significant fibrosis (FS2), but larger for detecting cirrhosis (FZ4). They showed good results especially for detecting cirrhosis. Although it is difficult to compare their results with ours because our pilot study cohort was small and had no severe cirrhosis, the competence of FibroScanw to detect

[1] Nord JH. Biopsy diagnosis of cirrhosis: blind percutaneous versus guided direct vision technique. A review. Gastrointest Endosc 1982; 28:102–104. [2] Bedossa P, Darge`re D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003;38:1449–1457. [3] Brunt EM. Grading and staging the histopathological lesions of chronic hepatitis: The Knodell histology activity index and beyond. Hepatology 2000;31:241–246. [4] Hu¨bscher SG. Histological grading and staging in chronic hepatitis: clinical applications and problems. J Hepatol 1998;29:1015–1022. [5] Phal A, Behling C, Oliver D, Kilani M, Monson P, Hassanein T. Serum aminotransferase levels and platelet counts as predictors of degree of fibrosis in chronic hepatitis C virus infection. Am J Gastroenterol 2001;96:3053–3055. [6] Pilette C, Rousselet MC, Bedossa P, Chappard D, Oberti F, Rifflet H, et al. Histopathological evaluation of liver fibrosis: quantitative image analysis vs semi-quantitative scores. J Hepatol 1998;28:439–446. [7] Lu LG, Zeng MD, Wan MB, Li CZ, Mao YM, Zhou XQ, et al. Grading and staging of hepatic fibrosis, and its relationship with non invasive diagnostic parameters. World J Gastroenterol 2003;9: 2574–2578. [8] Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38:518–526. [9] Taura T, Nakamura K, Takashima S, Kaminou T, Yamada R, Shuto T, et al. Heterogeneity of hepatic parenchymal enhancement on computed tomography during arterial portography: quantitative analysis of correlation with severity of hepatic fibrosis. Hepatol Res 2001;20:182–192. [10] Zuckerman E, Slobodin G, Sabo E, Yeshurun D, Naschitz JE, Groshar D. Quantitative liver-spleen scan using single photon emission computerized tomography (SPECT) for assessment of hepatic function in cirrhotic patients. J Hepatol 2003;39:326–332. [11] Colli A, Fraquelli M, Andreoletti M, Marino B, Zuccoli E, Conte D. Severe liver fibrosis or cirrhosis: Accuracy of US for detectionAnalysis of 300 cases. Radiology 2003;227:89–94. [12] Aube´ C, Oberti F, Krali N, Namour MA, Loisel D, Tanguy JY, et al. Ultrasonographic diagnosis of hepatic fibrosis or cirrhosis. J Hepatol 1999;30:472–478. [13] Badawi AM, Derbala AS, Youssef ABM. Fuzzy logic algorithm for quantitative tissue characterization of diffuse liver diseases from ultrasound images. Int J Med Inform 1999;55:135–147. [14] Sandrin L, Fourquet B, Hasquenoph JM, Yon S, Fournier C, Mal F, et al. Transient elastography: A new noninvasive method for assessment of hepatic fibrosis. Ultrasound Med Biol 2003;29: 1705–1713. [15] Wang Y, Itou K, Taniguchi N, Toei H, Kawai F, Nakamura M, et al. Studies on tissue characterization by texture analysis with cooccurrence matrix method using ultrasonography and CT imaging. J Med Ultrasonics 2002;29:211–223.

H. Yamada et al. / Journal of Hepatology 44 (2006) 68–75 [16] Yeh WC, Huang SW, Li PC. Liver fibrosis grade classification with B-mode ultrasound. Ultrasound Med Biol 2003;29:1229–1235. [17] Hung CH, Lu SN, Wang JH, Lee CM, Chen TM, Tung HD, et al. Correlation between ultrasonographic and pathologic diagnoses of hepatitis B and C virus-related cirrhosis. J Gastroenterol 2003;38: 153–157. [18] Hirata M, Akbar SMF, Horiike N, Onji M. Noninvasive diagnosis of the degree of hepatic fibrosis using ultrasonography in patients with chronic liver disease due to hepatitis C virus. Eur J Clin Invest 2001; 31:528–535. [19] Oosterveld BJ, Thijssen JM, Hartman PC, Rosenbusch GJ. Detection of diffuse liver disease by quantitative echography: Dependence on A priori choice of parameters. Ultrasound Med Biol 1993;19:21–25. [20] Zheng RQ, Wang QH, Lu MD, Xie SB, Ren J, Su ZZ. Liver fibrosis in chronic viral hepatitis: An ultrasonographic study. World J Gastroenterol 2003;9:2484–2489. [21] Yeh WC, Li PC, Jeng YM, Hsu HC, Kuo PL, Li ML. Elastic modulus measurements of human liver and correlation with pathology. Ultrasound Med Biol 2002;28:467–474. [22] Fukuda H, Ebara M, Kobayashi A, Sugiura N, Yoshikawa M, Saisho H, et al. An image analyzing system using an artificial neural network for evaluating the parenchymal echo pattern of cirrhotic liver and chronic hepatitis. IEEE Trans Biomed Eng 1998; 45:396–400. [23] Yamaguchi T, Hachiya H, Kamiyama N, Moriyasu F. Examination of the spatial correlation of statistics information in the ultrasonic echo from diseased liver. Jpn J Appl Phys 2002;41:3585–3589.

75

[24] Yamaguchi T, Hachiya H, Kamiyama N, Ikeda K, Moriyasu F. Estimation of characteristics of echo envelope using RF echo signal from the liver. Jpn J Appl Phys 2001;40:3900–3904. [25] Desmet VJ, Gerber M, Hoofnagle JH, Manns M, Scheuer PJ. Classification of chronic hepatitis: Diagnosis, grading and staging. Hepatology 1994;19:1513–1520. [26] Masseroli M, Caballero T, O’Valle F, Del Moral RM, Pe´rezMilena A, Del Moral RG. Automatic quantification of liver fibrosis: design and validation of a new image analysis method: comparison with semi-quantitative indexes of fibrosis. J Hepatol 2000;32: 453–464. [27] Wright M, Thursz M, Pullen R, Thomas H, Goldin R. Quantitative versus morphological assessment of liver fibrosis: semi-quantitative scores are more robust than digital image fibrosis area estimation. Liver Int 2003;23:28–34. [28] Lin XZ, Horng MH, Sun YN, Shiesh SC, Chow NH, Guo XZ. Computer morphometry for quantitative measurement of liver fibrosis: Comparison with Knodell’s score, colorimetry and conventional description reports. J Gastroenterol Hepatol 1998; 13:75–80. [29] Downey DB, Fenster A, Williams TC. Clinical utility of threedimensional US. RadioGraphics 2000;20:559–571. [30] Ziol M, Handra-Luca A, Kettaneh A, Chirstidis C, Mal F, Kazemi F, et al. Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C. Hepatology 2005;41:48–54. [31] Caste´ra L, Vergnoil J, Foucher J, Le Bail B, Chanteloup E, Hasser M, et al. Prospective comparison of transient elastography, fibrotest, APRI, and liver biopsy for assessment of fibrosis in chronic hepatitis C. Gastroenterology 2005;128:343–350.