Automatic counting of immunocytochemically stained cells

In this work is described the development of an automatic color image segmentation and cell counting system for immunocytochemical analysis of stained tissue samples. The system is designed to automatically count the total number of positive and negative cells in tissue samples treated with cytokines DNA probes of pigs naturally parasitized with Taenia solium metacestodes and using in situ hybridization. The objectives of automatic counting are to improve the reproducibility of the analysis and the processing time of large image batches. A Bayes classifier was used for color image segmentation. Improved watershed segmentation combined with edge detection was used to isolate individual cells which are automatically labeled from the color segmented images. Preliminary results of 174 digital images are reported. The correlation coefficient of the automatic system with manual counting is 0.8 approx. Processing of each digital image takes 20 s on a SUN BLADE 2000.