BACKGROUND: There is growing interest in the role of hepatic steatosis in liver injury. The current standard for steatosis assessment is histological grading, although there is variability in the scoring systems used. AIMS: The aim of this study was to compare steatosis assessment by image analysis and histological grading. METHODS: Three methods were used to measure steatosis: histological grading (from 0 to 4); estimation of the percentage of hepatocytes (to nearest 5%) with steatosis; and computer-assisted image analysis. Image analysis was performed on multiple fields for each biopsy with image pro plus 4.5, with steatotic droplets identified on the basis of shape, colour and size. Computer-selected objects were reviewed to ensure that these were steatotic droplets. The predictive accuracy of the three techniques was assessed using measures of obesity and insulin resistance (homeostasis model assessment) as the outcome variables. RESULTS: There was a strong correlation between the results of image analysis and histological grade (r(s)=0.89, P<0.01), and estimated per cent steatosis (r(s)=0.93, P<0.01). The variability in the area of steatosis calculated by image analysis in different fields of a biopsy correlated with the total steatosis area (r(s)=0.93, P<0.01). CONCLUSIONS: Image analysis did not offer any additional predictive value when the association between degree of obesity or insulin resistance was correlated with the different methods of assessing steatosis. Image analysis allows measurement of area of steatosis in liver biopsy material and generates a continuous variable that facilitates statistical analysis. These aspects may prove beneficial in research settings.
BACKGROUND: There is growing interest in the role of hepatic steatosis in liver injury. The current standard for steatosis assessment is histological grading, although there is variability in the scoring systems used. AIMS: The aim of this study was to compare steatosis assessment by image analysis and histological grading. METHODS: Three methods were used to measure steatosis: histological grading (from 0 to 4); estimation of the percentage of hepatocytes (to nearest 5%) with steatosis; and computer-assisted image analysis. Image analysis was performed on multiple fields for each biopsy with image pro plus 4.5, with steatotic droplets identified on the basis of shape, colour and size. Computer-selected objects were reviewed to ensure that these were steatotic droplets. The predictive accuracy of the three techniques was assessed using measures of obesity and insulin resistance (homeostasis model assessment) as the outcome variables. RESULTS: There was a strong correlation between the results of image analysis and histological grade (r(s)=0.89, P<0.01), and estimated per cent steatosis (r(s)=0.93, P<0.01). The variability in the area of steatosis calculated by image analysis in different fields of a biopsy correlated with the total steatosis area (r(s)=0.93, P<0.01). CONCLUSIONS: Image analysis did not offer any additional predictive value when the association between degree of obesity or insulin resistance was correlated with the different methods of assessing steatosis. Image analysis allows measurement of area of steatosis in liver biopsy material and generates a continuous variable that facilitates statistical analysis. These aspects may prove beneficial in research settings.
Authors: Sekou R Rawlins; Ola El-Zammar; J Michael Zinkievich; Nancy Newman; Robert A Levine Journal: Dig Dis Sci Date: 2010-05-12 Impact factor: 3.199