Computer identification of degenerated urothelial cells.

Computer algorithms were developed to identify very poorly preserved, degenerated urothelial cells from the urinary sediment. Cell images were analyzed by two methods: the first one based on nucleus-finding algorithms (determination of nuclear boundaries) and the second based on the analysis of histograms of optical density. Discrimination between benign and malignant degenerated cells could be accomplished by both methods, although the misclassification of cells rose from +/- 10% by the first approach to +/- 15% by the second approach. Most important, however, useful algorithms were obtained and incorporated into the TICAS file for the identification of this large group of urothelial cells as a significant step in the automated analysis of the cells in thr urinary sediment.