Hepatitis C is a common viral infection of the liver. The degree of inflammation associated with the infection is normally estimated manually from a liver biopsy, by considering the quantity and nature of inflammatory cells. This paper presents an automated pattern recognition system for the quantification of inflammatory cells in liver biopsies. Initially, images are corrected for colour variation. Features are then extracted for from colour biopsy images at positions of interest identified by adaptive thresholding and clump decomposition. A sequential floating search method and principal component analysis are used to reduce the dimensionality of the feature vector. Manually annotated training images allow supervised training by providing the class membership for each position of interest. Gaussian parametric and Gaussian mixture model density estimation methods are compared and are used to classify cells as either inflammatory or healthy via Bayes' theorem. The system is optimised using a response surface method, where the response or system performance is derived from the area under the receiver operating characteristic curve. The optimised system is then tested on test images previously ranked by a number of observers with varing levels of pathology experience. The observers results are compared to the automated system using Spearman rank correlation. Results show that this system can rank 15 test images, with varying degrees of inflammation, in strong agreement with five expert pathologists.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)