March 11, 2014

An Inexpensive and Worldwide Available Digital Image Analysis Technique for Histological Fibrosis Quantification in Chronic Hepatitis C

Journal of Viral Hepatitis

C. F. F. Campos, D. D. Paiva, H. Perazzo, P. S. Moreira, L. F. F. Areco, C. Terra, R. Perez, F. A. F. Figueiredo

J Viral Hepat. 2014;21(3):216-222.

Abstract and Introduction

Abstract

Hepatic fibrosis staging is based on semiquantitative scores. Digital imaging analysis (DIA) appears more accurate because fibrosis is quantified in a continuous scale. However, high cost, lack of standardization and worldwide unavailability restrict its use in clinical practice. We developed an inexpensive and widely available DIA technique for fibrosis quantification in hepatitis C, and here, we evaluate its reproducibility and correlation with semiquantitative scores, and determine the fibrosis percentage associated with septal fibrosis and cirrhosis. 282 needle biopsies staged by Ishak and METAVIR scores were included. Images of trichrome-stained sections were captured and processed using Adobe® Photoshop® CS3 and Adobe® Bridge® softwares. The percentage of fibrosis (fibrosis index) was determined by the ratio between the fibrosis area and the total sample area, expressed in pixels calculated in an automated way. An excellent correlation between DIA fibrosis index and Ishak and METAVIR scores was observed (Spearman's r = 0.95 and 0.92; P < 0.001, respectively). Excellent intra-observer reproducibility was observed in a randomly chosen subset of 39 biopsies with an intraclass correlation index of 0.99 (95% CI, 0.95–0.99). The best cut-offs associated with septal fibrosis and cirrhosis were 6% (AUROC 0.97, 95% CI, 0.95–0.99) and 27% (AUROC 1.0, 95% CI, 0.99–1), respectively. This new DIA technique had high correlation with semiquantitative scores in hepatitis C. This method is reproducible, inexpensive and available worldwide allowing its use in clinical practice. The incorporation of DIA technique provides a more complete evaluation of fibrosis adding the quantification to architectural patterns.

Introduction

Chronic hepatitis C is a major public health problem and continues to be a leading cause of chronic liver disease and cirrhosis worldwide.[1] Liver fibrosis staging is still essential in management of these patients. The accurate assessment of hepatic fibrosis plays a critical role determining antiviral treatment, screening strategies and prognosis.[1,2] Furthermore, patients with advanced fibrosis and cirrhosis have a poor prognosis, presenting lower survival rate than mild fibrosis patients.[3–8]

Currently, liver biopsy is still considered the gold standard for fibrosis staging.[9] Traditional histological assessment is based on semiquantitative scoring systems (Ishak and METAVIR score).[10,11] Although frequently used, these scoring systems do not allow a precise quantification of liver fibrosis. Moreover, they are subjective measurements with high rates of intra- and interobserver variability.[12]

During the last years, methods to quantify liver fibrosis through computer software, called digital image analysis (DIA), have been developed.[13–22] DIA allows quantitative assessment of fibrosis using pixel counting to estimate fibrosis area. The great advantage of this method over semiquantitative scores is that DIA is a truly quantitative method, not influenced by subjective visual interpretation of the observer.[23] This new technology has been applied to estimate liver fibrosis in chronic viral hepatitis.[21,24] However, the use of different softwares, high cost, lack of standardization of this method and worldwide unavailability restrained its use to a few numbers of specialized centres.[25]

The primary aim of this study was to develop an inexpensive and worldwide available DIA technique for fibrosis quantification in liver biopsies of patients with chronic hepatitis C. Secondary aims were (i) to compare METAVIR and Ishak scoring systems with this new quantitative assessment of fibrosis, (ii) to determine the intra-observer reproducibility of the fibrosis index and (iii) to establish the most accurately cut-off associated to septal fibrosis and cirrhosis.

Material and Methods

This prospective observational study included liver core needle biopsies obtained from 282 patients with chronic hepatitis C at the Pathology Department, University of the State of Rio de Janeiro, Brazil. Hepatitis C infection was characterized by the presence of HCV-RNA in blood serum, and liver biopsy was performed according to routine clinical practice.

The study protocol was conducted in accordance with the ethical principles, the guidance of the Helsinki Declaration and the Good Clinical Practice Guidelines. The study was approved by the local Ethics Committee and all patients signed the informed consent.

Histological Analysis

All samples were obtained with 14G Menghini needles, fixed in a 10% neutral-buffered formalin solution and cut in 5-mm-thick sections. Routinely, haematoxylin and eosin, Masson's trichrome and reticulin stains were performed. The inclusion criteria of the liver sample were at least five complete portal tracts and size of 15 mm.

A single experienced pathologist (P.S.M.) evaluated all samples, blinded to clinical data and DIA results. Staging was carried out using the Ishak and METAVIR scores.[10,11] Based on these two semiquantitative scores, two clinical scenarios were considered as follows: septal fibrosis (Ishak score ≥ 3 or METAVIR ≥ 2), indicative of antiviral treatment and cirrhosis (Ishak score 6 or METAVIR 4), indicative of varices and hepatocellular carcinoma surveillance.

Digital Image Analysis

The digital imaging acquisition system consisted of an Olympus E-330 7.5 megapixels camera (Olympus Corporation, Tokyo, Japan) attached by an adapter to a trinocular Olympus BX 41 microscope (Olympus Corporation). The camera was connected to an analogue Sony PVM-14N5U Trinitron monitor (Olympus Corporation), to facilitate the visualization of the microscopic fields. The overall costs of this assemble were approximately US$ 5650.

Each sample was digitalized in a sequential way, respecting the linear distribution of microscopic fields of the core needle biopsies, using a 40× magnification objective. The images were captured using automatic adjustments for white and luminosity of the camera, with maximal resolution (3748 × 2736 pixels), and saved as 24 bits RGB images in the Joint Photographic Experts Group (JPEG) format. Depending on the size of the sample, six to eighteen images were captured and saved in folders identified with each protocol number, in a personal computer (PC). Using Adobe® Bridge® (Adobe Corporation, San Jose, CA, USA), running in a Windows® 7 64 bits environment, the folder containing the captured images of each individual biopsy was identified. The average cost of the software was U$ 1200.

The images were selected and the following command line was taken as follows: 'Tools', 'Photoshop' and 'Photomerge'. Following this command line, the Adobe® Photoshop® program started automatically and a menu box containing the instructions with the selected files was displayed. In the checkboxes, the options 'Auto' in the 'Layout' space and 'Blend images together' in 'Source files' were, by default, selected. Afterwards, running the command, the program automatically mixed the digitalized microscopic fields and a panoramic wide field image representing the entire area of the biopsy stained with Masson's trichrome was constructed. This wide field panoramic image was saved in the JPEG format, with maximal resolution, representing, each one, a single, totally digitalized virtual image of the biopsy.

The following actions were taken for individual adjustments of the wide field biopsy images: extraction of the background and undesired elements like fragments of the capsule, soft tissues and thick vessels walls, using the command 'Extract' in the 'Filters' tool menu. Using the 'Auto levels' action, in the 'Image', 'Adjustments' menu, lightness and brightness were automatically adjusted for each wide field image. Afterwards, following the technique described by Dahab et al.,[20] using the 'Selective Color' command presented under the 'Image' drop-down menu, the red, magenta, cyan and blue colours were adjusted after checking the 'Relative' option in the 'Selective Color' dialogue box. From the colour table, the red and magenta colours were chosen sequentially and their cyan component was reduced to -100% and the magenta component expanded to +100%. The cyan and blue colours were selected, their cyan component expanded to +100% and their magenta component reduced to –100%.

For the automatic calculation of the pixels corresponding to the total area of the sample, using the 'Magic Wand Tool', from the tools palette, the extracted background was selected and, using the command 'Select Inverse' of the 'Magic Wand Tool', the area corresponding to the biopsy tissue stained in cyan/blue, red/magenta and grey was, then, selected. The histogram command serves as an internal measurement of tonal distribution as the basis for automated image manipulation. The histogram of the selected area, which represented the totality of pixels of the sample counted the total area of the digitalized biopsy in pixels. Next, the 'Magic Wand Tool' and the 'Similar' command were applied again and all the cyan/blue area of the sample, corresponding to the fibrous tissue, was selected. The histogram of the selected area represented the totality of fibrous tissue in the sample.

The mean time spent in the process of acquisition and image analysis was around 15 min. Smaller samples (consisting of six images before the generation of the wide field image) took around 10 min from the capture process until the last steps of image analysis, while larger samples (consisting of eighteen images before the generation of the wide field image) consumed around 20 min.

The result of this process is shown in Fig. 1. The fibrosis index (FI) was the total area of fibrosis divided by the total area of the section multiplied by 100, as showed by Dahab et al.:[20]

820294-fig1

Figure 1. Wide field image representing the entire biopsy (Ishak score 1, METAVIR F1). Fibrosis area = 2800 pixels. Total area = 122495 pixels. Fibrosis index = 2800 × 100/122495 = 2.2%.

Intra-observer reproducibility was assessed in a randomly chosen a subset of 39 biopsies. The pathologist was blinded to the initial measurements and to the Ishak and METAVIR scores.

Statistical Analysis

Statistical analyses were performed using SPSS (v.17.0.0) software (SPSS Inc., Chicago, IL, USA) and MedCalc software package version 12.2 (MedCalc Software, Mariakerke, Belgium). In all analyses, significance was determined when P < 0.05 assuming two-tailed tests. Spearman's correlation index was used to evaluate the correlation of the DIA technique with METAVIR and Ishak stages. The intraclass correlation coefficient was used to measure the intra-observer reproducibility. The receiver operating characteristic (ROC) curve and area under the ROC (AUROC) curve were applied to establish the fibrosis index associated with septal fibrosis and cirrhosis.

Results

This study showed that was possible to develop an inexpensive and worldwide DIA technique for liver fibrosis quantification using hardware and software easily found in the market. The total cost of the system was less than US$ 8000.

Correlation Between DIA Technique and Semiquantitative Scores

The fibrosis indexes obtained by DIA according to Ishak and METAVIR scores are shown in Table 1 and in Figs 2 & 3, respectively. An excellent correlation between FI obtained by DIA and Ishak and METAVIR semiquantitative scoring systems was observed (Spearman's r = 0.95 and 0.92, respectively, P < 0.001).

Table 1.  Fibrosis index according to staging by Ishak and METAVIR scores

Ishak score METAVIR score
Stage (n) Fibrosis index (%) Stage (n) Fibrosis index (%)
0 (5) 0.8 ± 0.05 0 (5) 0.8 ± 0.05
1 (41) 2.5 ± 0.7 1 (99) 3.9 ± 1.7
2 (58) 4.9 ± 1.5 2 (79) 7.4 ± 1.5
3 (79) 4.1 ± 1.5 3 (77) 20.4 ± 5
4 (28) 15.3 ± 3.5 4 (22) 34.6 ± 1.6
5 (49) 23.3 ± 2.9
6 (22) 34.6 ± 1.6

820294-fig2

Figure 2. Distribution of mean (± confidence internal (CI)) fibrosis percentage data by Ishak scores.

820294-fig3

Figure 3. Distribution of mean (± confidence internal (CI)) fibrosis percentage data by METAVIR scores.

Assessment of Intra-observer Reproducibility

To evaluate the reproducibility of this new technique, a subset of 39 randomly specimens were re-evaluated by the same observer. These samples were representative of all stages of fibrosis, and the observer was blinded to clinical data and first measurements results. Excellent intra-observer reproducibility was observed, resulting in an intraclass correlation index of 0.99 (95% CI 0.95–0.99).

Diagnosis of Septal Fibrosis and Cirrhosis

ROC analysis for fibrosis index yielded an optimum cut-off of 6% with an AUROC of 0.97 (95% CI, 0.95–0.99) as the best power distinction for septal fibrosis according to both semiquantitative scoring systems providing a sensitivity of 91% (95% CI, 86–95%) and specificity of 99% (95% CI, 95–100%). Furthermore, the most accurate cut-off for diagnosis of cirrhosis was 27% with an AUROC of 1.0 (95% CI, 0.99–1) according to both scores classification, presenting a sensitivity of 100% (95% CI, 85–100%) and specificity of 100% (95% CI, 99–100%).

Discussion

Liver fibrosis is a major parameter guiding the diagnosis and prognosis of chronic liver disease.[3] Studies based on liver biopsies to evaluate fibrosis in chronic hepatitis C usually relies on categorical scoring systems rather than direct measurement of the amount of fibrous tissue.[22] Several DIA techniques for liver fibrosis quantification have been developed.[13–22] It is considered as a promising tool because it provides results on a continuous scale, rather than merely five or seven qualitative stages.[22] In this article, we described an inexpensive and worldwide available DIA technique that provides objective quantification of fibrosis in liver biopsies without the need of expensive digital glass-scanning devices and third party software.

Adobe® Photoshop® is largely used in the biomedical literature and considered 'inexpensive and commonly available' imaging software.[19,20,26–29] Two previous studies described DIA techniques for liver fibrosis quantification using older versions of this software.[19,20] The newer versions can be easily acquired from the World Wide Web. This software upgrade combined with the improvement of hardware performance allowed the development of very powerful image analysis tools. These make possible the whole sample digitalization (virtual large field images of the biopsies) without needing expensive glass-scanning devices.

Although Sirius Red is known as one of the best techniques for collagen histochemistry quantification of liver fibrosis,[30] the authors preferred to use Masson's trichrome stain. This staining method is worldwide available in most pathology laboratories, and thus, more suitable to the purposes of the present study. A perfect contrast between fibrous tissue stained in blue and parenchyma stained in red was obtained following the 'Selective color' procedure described in the methodology, using the Masson's trichrome stain. We reinforce that similar results can be reached in a liver biopsy slide stained with Sirius Red by adjusting different colour pallets: reds and yellows. In samples stained by this histochemical method, red and yellow colours will be representative of fibrous tissue and parenchyma, respectively.

Our DIA technique had an excellent correlation with the traditional semiquantitative Ishak score and METAVIR classification (r = 0.95 and r = 0.92, respectively). Other studies have already observed a high correlation between DIA techniques and staging.[18,21] It demonstrated a high capacity of DIA techniques in discriminating the different stages of fibrosis. Our study includes a larger sample and evaluated a cheaper method. Another advantage of the DIA method described in this study is its reproducibility with high rates of intra-observer concordance. We showed an intraclass correlation index of 0.99 (95% CI 0.95–0.99). This reinforces its reliability. Further studies are welcome to define interobserver variability and intercenter reproducibility. A small amount of training time and minimal knowledge in informatics are enough to accurately reproduce the method described. We anticipate that an easy learning curve, low cost and use of worldwide available technology will permit the application of this method in most pathology laboratories.

Important clinical cut-off points for septal fibrosis and for cirrhosis were established in this study. Patients with fibrosis index higher than 6% would have indication for antiviral therapy, while patients with fibrosis index higher than 27% would have indication for endoscopy and hepatocellular carcinoma surveillance. The role of these cut-off points in clinical practice should be better investigated in further studies.

The strengths of our DIA method were that images were processed on colour RGB model, the area of fibrosis was not manually manipulated, the software has adjustments of colours allowing a greater contrast between fibrosis and normal hepatic tissue, and the hardware/software packet is inexpensive and worldwide available. It also should be highlighted that all the fibrosis stages were well represented in this large sample size. Thus, this DIA method can be apply in current clinical practice with a very low cost.

In our methodology, the quantification of fibrosis automatically selects all blue sample allowing to map fibrous tissue including the delicate perisinusoidal fibrosis, which is usually undetected by examiner's eye. The precisely quantification of perisinusoidal fibrosis might be an advantage of this method in quantitative assessment of liver fibrosis in patients with others chronic liver disease than CHC, especially nonalcoholic fatty liver disease (NAFLD). Digital quantification of fibrosis by this method can be a future tool to ease the differentiation between patients with simple steatosis from those with nonalcoholic steatohepatitis (NASH). It can also contribute for a more precise staging in the last condition.

The histological diagnosis based on semiquantitative architectural changes in association with DIA techniques may express more precisely the actual fibrosis state. This combination allows evaluating not only the fibrosis distribution but also the fibrosis amount. We agree with the idea that DIA techniques are not suitable for replacing traditional qualitative and semiquantitative liver biopsy evaluation.[31] More important, it should be considered as a complementary tool for the traditional histological methods.

Some potential applications for DIA techniques as complementary tool could be expected. In patients undergoing sequential liver biopsies showing the same stage, the variation in fibrosis index could estimate more precisely the progression of fibrosis. Another potential application could be in clinical trials for measuring the effect of novel antifibrogenic therapies when more objective and broader scale information is needed. This would be important for the validation of new noninvasive imaging techniques and indirect serum markers of fibrosis.[32,33] It could also be useful in case of discordance of staging among different pathologists.

In summary, the DIA technique for liver fibrosis quantification described in this article is inexpensive and was developed using worldwide available technologies. The process is simple, reproducible and showed a high correlation with semiquantitative scores. It provides a more complete evaluation of fibrosis adding the quantification to the architectural patterns.

References

1.EASL. Clinical Practice Guidelines management of hepatitis C virus infection. J Hepatol 2011; 55(2): 245–264.

2.Ghany MG, Strader DB, Thomas DL, Seeff LB. Diagnosis, management, and treatment of hepatitis C: an update. Hepatology 2009; 49(4): 1335–1374.

3.Khan MH, Farrell GC, Byth K et al. Which patients with hepatitis C develop liver complications? Hepatology 2000; 31(2): 513–520.

4.Ngo Y, Munteanu M, Messous D et al. A prospective analysis of the prognostic value of biomarkers (FibroTest) in patients with chronic hepatitis C. Clin Chem 2006; 52(10): 1887–1896.

5.Poynard T, Bedossa P, Opolon P. Natural history of liver fibrosis progression in patients with chronic hepatitis C. The OBSVIRC, METAVIR, CLINIVIR, and DOSVIRC groups. Lancet 1997; 349(9055): 825–832.

6.Kobayashi M, Tanaka E, Sodeyama T, Urushihara A, Matsumoto A, Kiyosawa K. The natural course of chronic hepatitis C: a comparison between patients with genotypes 1 and 2 hepatitis C viruses. Hepatology 1996; 23(4): 695–699.

7.Shev S, Dhillon AP, Lindh M et al. The importance of cofactors in the histologic progression of minimal and mild chronic hepatitis C. Liver 1997; 17(5): 215–223.

8.Vergniol J, Foucher J, Terrebonne E et al. Noninvasive tests for fibrosis and liver stiffness predict 5-year outcomes of patients with chronic hepatitis C. Gastroenterology 2011; 140 (7): 1970–1979, 9 e1–3.

9.Castera L, Bedossa P. How to assess liver fibrosis in chronic hepatitis C: serum markers or transient elastography vs. liver biopsy? Liver Int 2011; 31(Suppl 1): 13–17.

10.Ishak K, Baptista A, Bianchi L et al. Histological grading and staging of chronic hepatitis. J Hepatol 1995; 22(6): 696–699.

11.Intraobserver and interobserver variations in liver biopsy interpret tation in patients with chronic hepatitis C The French METAVIR Cooperative Study Group. Hepatology 1994; 20(1 Pt 1): 15–20.

12.Cadranel JF, Rufat P, Degos F. Practices of liver biopsy in France: results of a prospective nationwide survey. For the Group of Epidemiology of the French Association for the Study of the Liver (AFEF). Hepatology 2000; 32(3): 477–481.

13.Manabe N, Chevallier M, Chossegros P et al. Interferon-alpha 2b therapy reduces liver fibrosis in chronic non-A, non-B hepatitis: a quantitative histological evaluation. Hepatology 1993; 18(6): 1344–1349.

14.Chevallier M, Guerret S, Chossegros P, Gerard F, Grimaud JA. A histological semiquantitative scoring system for evaluation of hepatic fibrosis in needle liver biopsy specimens: comparison with morphometric studies. Hepatology 1994; 20(2): 349–355.

15.Kage M, Shimamatu K, Nakashima E, Kojiro M, Inoue O, Yano M. Long-term evolution of fibrosis from chronic hepatitis to cirrhosis in patients with hepatitis C: morphometric analysis of repeated biopsies. Hepatology 1997; 25(4): 1028–1031.

16.Pilette C, Rousselet MC, Bedossa P et al. Histopathological evaluation of liver fibrosis: quantitative image analysis vs semi-quantitative scores Comparison with serum markers. J Hepatol 1998; 28(3): 439–446.

17.O'Brien MJ, Keating NM, Elderiny S et al. An assessment of digital image analysis to measure fibrosis in liver biopsy specimens of patients with chronic hepatitis C. Am J Clin Pathol 2000; 114(5): 712–718.

18.Masseroli M, Caballero T, O'Valle F, Del Moral RM, Perez-Milena 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(3): 453–464.

19.Arima M, Terao H, Kashima K, Arita T, Nasu M, Nishizono A. Regression of liver fibrosis in cases of chronic liver disease type C: quantitative evaluation by using computed image analysis. Intern Med 2004; 43(10): 902–910.

20.Dahab GM, Kheriza MM, El-Beltagi HM, Fouda AM, El-Din OA. Digital quantification of fibrosis in liver biopsy sections: description of a new method by Photoshop software. JGastroenterol Hepatol 2004; 19(1): 78–85.

21.Lazzarini AL, Levine RA, Ploutz-Snyder RJ, Sanderson SO. Advances in digital quantification technique enhance discrimination between mild and advanced liver fibrosis in chronic hepatitis C. Liver Int 2005; 25(6): 1142–1149.

22.Goodman ZD, Becker RL Jr, Pockros PJ, Afdhal NH. Progression of fibrosis in advanced chronic hepatitis C: evaluation by morphometric image analysis. Hepatology 2007; 45(4): 886–894.

23.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. LiverInt 2003; 23(1): 28–34.

24.Xie SB, Ma C, Lin CS, Zhang Y, Zhu JY, Ke WM. Collagen proportionate area of liver tissue determined by digital image analysis in patients with HBV-related decompensated cirrhosis. HepatobiliaryPancreat Dis Int 2011; 10(5): 497–501.

25.Rosai J. Rosai and Ackerman's Surgical Pathology. 10a ed edn, New York, NY: Elsevier 2011.

26.Gatlin CL, Schaberg ES, Jordan WH, Kuyatt BL, Smith WC. Point counting on the Macintosh. A semiautomated image analysis technique. Anal Quant Cytol Histol 1993; 15 (5): 345–350.

27.Brunner J, Krummenauer F, Lehr HA. Quantification of video-taped images in microcirculation research using inexpensive imaging software (Adobe Photoshop). Microcirculation 2000; 7(2): 103–107.

28.Marchevsky AM, Wan Y, Thomas P, Krishnan L, Evans-Simon H, Haber H. Virtual microscopy as a tool for proficiency testing in cytopathology: a model using multiple digital images of Papanicolaou tests. Arch Pathol Lab Med 2003; 127 (10): 1320–1324.

29.Denissen H, Dozic A. Photometric assessment of tooth color using commonly available software. EurJ Esthet Dent 2010; 5(2): 204–215.

30.Standish RA, Cholongitas E, Dhillon A, Burroughs AK, Dhillon AP. An appraisal of the histopathological assessment of liver fibrosis. Gut 2006; 55: 569–578.

31.Scheuer PJ. Liver biopsy size matters in chronic hepatitis: bigger is better. Hepatology 2003; 38(6): 1356–1358.

32.Poynard T, Ngo Y, Munteanu M et al. Biomarkers of liver injury for hepatitis clinical trials: a meta-analysis of longitudinal studies. AntivirTher 2010; 15(4): 617–631.

33.Degos F, Perez P, Roche B et al. Diagnostic accuracy of FibroScan and comparison to liver fibrosis biomarkers in chronic viral hepatitis: a multicenter prospective study (the FIBROSTIC study). J Hepatol 2010; 53(6): 1013–1021.

Source

No comments:

Post a Comment