Liver Fibrosis Quantitation Via Image Analysis: Correlation with Pathologist Assessment and Clinical Parameters
Jenny Mas-Moya, Pelin Bagci, Kee-Taek Jang, Michael J Lee, Hiroumi D Kitajima, Diego R Martin, N Volkan Adsay, Alton B Farris. Emory University, Atlanta, GA; University of Arizona, Tucson, AZ
Background: The assessment of hepatic fibrosis is susceptible to sampling and pathologist observer variability, sometimes making consistent, objective fibrosis severity appraisal difficult. Image analysis tools have been used with varying success; however, technology limitations and lack of correlation with clinical parameters have prevented the application of potentially more objective image analysis methods.
Design: Transplant hepatectomy specimens were sectioned in the axial plane, and at least 5 sections were taken for histology from each hepatectomy. Whole slide images (WSIs) of trichrome-stained sections were obtained using a whole slide scanner. In sections containing the least and greatest fibrosis, %fibrosis was quantitated using a positive pixel count (PPC) algorithm tuned to detect fibrosis; and an average PPC measurement was obtained. Image analysis measurements were compared with pathologist assessments of %fibrosis and clinical parameters.
Results: All cases (n=46) were cirrhotic hepatectomies in patients with end stage liver disease from a variety of disorders. A wide range of %fibrosis was present (average±S.D. = 33.0±16.6 and 30.6±16.4 as determined by visual and PPC methods, respectively). There was a statistically significant correlation between the visual and PPC methods (r=0.58, p<0.0001). The visual and PPC methods also showed correlation with MELD score, PT, and PTT (r=0.37 and 0.36, 0.37 and 0.50, and 0.34 and 0.42, respectively with p = 0.02 to <0.0001). The relationships between the visual and PCC methods and total bilirubin, AST, ALT, GGT, and albumin were not statistically significant.
Conclusions: A wide range of %fibrosis was observed according to image analysis and visual assessment, and these measurement methods correlated with each other and some clinical parameters. These findings suggest that image analysis methods have the potential to provide an objective assessment of hepatic fibrosis. Future improvement in image analysis methods could improve correlation with clinical parameters, assisting pathologists in providing the best possible appraisal of hepatic fibrosis.
Wednesday, March 6, 2013 1:00 PM
Poster Session VI # 114, Wednesday Afternoon