Computer-aided analysis of nuclear stained breast cancer cell images

This paper describes a computer-aided system for analyzing stained breast cancer cell images. The procedure for the analysis approach is composed of three steps. First, the cancer cells in the microscopic image are segmented based on neural network and mathematical morphology. Next, the features consisting of average values of L*; a*, b*, area and circularity ratio of each cell are extracted. Finally, the classification is operated using the Euclidean distance in CIELab color space. Results from our computer-aided analysis system show a promising solution to the traditional manual analysis. That is, the cancer cell is appropriately segmented. The classifications of segmented cell type based on the Euclidean distance in CIELab color space agree with visual inspection very well.