Prostatic intraepithelial neoplasia. Quantitation of the basal cell layer with machine vision system

The aim of this paper is to report on a fully automated procedure to quantify the epithelial cell components in prostatic intraepithelial neoplasia with particular emphasis on the basal cell layer (BCL). Scene segmentation was guided by a knowledge-based system of digitized images recorded from histological section immunostained against high molecular weight keratin and counterstained with hematoxylin and eosin. Scene segmentation involved two major stages. First, the system located and identified the duct and segmented the scene, resulting in "intermediate segmentation products." This stage was followed by a reconstruction process in which the segmentation products (i.e., the lumen and the darkly and lightly stained epithelial cell components) were assembled to form the microscopic structure to achieve working unity. Following this, histometric measurements were made of the reconstructed scene. Computed were the percentage of the duct contour with BCL (90%), and the number and length of gaps in the BCL (19, ranging from 10 to 90 microns). Automated analysis of the BCL is accurate and provides information not readily identified by human examination.