Automated scoring of liver fibrosis through combined features from different collagen groups

Abstract

Liver biopsy remains the gold standard in monitoring progression of liver fibrosis associated with an abnormal increase in collagen. Descriptive scoring systems are still being widely used to grade biopsy samples. In this study, we propose a new set of features by clustering collagen fibers into three groups first based on their localization and connectivity properties, and then by extracting morphological features of collagen fibers. The new feature set is compared to the earlier features used in classification of liver fibrosis, which were based on the total amount of collagen fibers. Our results show that new features lead to more accurate grading of liver fibrosis. © 2011 IEEE.