Romero, E.Cuisenaire, O.Moulin, P.Macq, B.2006-06-142006-06-142006-06-14200210.1109/MCTE.2002.1175066https://infoscience.epfl.ch/handle/20.500.14299/231349This paper presents a reliable, fast and efficient method for measuring the volume density of pancreatic endocrine volume density. The algorithm segments digitized images in three different classes: the endocrine (En), exocrine (Ex) and artifact (At) components. A statistical classifier baased on the k-Nearest Neighbour (k-NN) decision rule in the RGB color space was compared with a standard point counting technique. The k-NN rule classifies other pixels in the class that is mostly respresented among the k nearest training samples in the RGB space, which is efficiently implemented with a fast k-distance transform algorithm. All extracted areas were quantified in absolute (um2) and relative (%) values. The different tissues were point counting determined and their quantifications statistically compared with those obtained semi-automatically. All anayses were performed by an expert pathologist and showed no significant differences between the two approaches.LTS1Databases and Information Systemstext::conference output::conference proceedings::conference paper