Automatic segmentation for ovarian cancer immunohistochemical image based on chroma criterion

Immunohistochemical color image segmentation has important application value for quantitative assessment of immunohistochemical image. In this paper, an automatic segmentation method was proposed according to characteristics of color immunohistochemical images. First of all, we established a Chromatics criteria in RGB space so that positive cells regional and negative cells region were separated automatically and two new images-image A and image B were generated. Then, on the basis of the results, the improved ISODATA clustering algorithm was used in segmenting: (1) Determine the initial cluster center of image A and image B; (2) Extract positive cells of image A on R component and negative cells of image B on B component using ISODATA clustering algorithm. The improved ISODATA clustering algorithm reduced the sample amounts and enhances the computing speed. The experimental results showed that the method can be a good segmentation of ovarian cancer immunohistochemical images.