An Automatic Image Analysis System for Quantitative Assessment of Liver Fibrosis

Abstract

Liver biopsy is the standard for the evaluation of liver diseases. The severity of fibrosis obtained by biopsy test is considered to reflect the stage of chronic liver disease. However, the current biopsy reading remains quite subjective, or at best semiquantitative. Therefore, developing an objective and reliable image analysis system for evaluating the liver fibrosis will be a major advance in the diagnosis and staging of chronic liver disease. In this paper, we develop an automatic image analysis syatem, which consists of a microscope, a computer-driven slide-driver, and the software for image acquisition, processing and data analysis. Some image analysis procedures, including color model selection, histogram-based normalization, clustering, moment-persevering thresholding and ranking filter, have been developed and employed in this system for tissue characterization. In addition, the computerized motor driver and the x-y directional stage to move specimens, are designed and installed on an optical microscope to construct a fully automated system. The system is capable of computing the percentage of fibrous area to the whole liver tissue area as an index, called computer morphometry (CM) csore to reflect the severity of liver fibrosis. Experimental results show that the proposed method correlates well these methods as expected, however, is more stable and quantitative than these conventional methods.