A new method for automated segmentation of double stained immunohistochemical images

MENG Jun,LU Wei,WANG Li-zhen,YU Zhi-wei,GUO Yan,CHENG Ming,SHAO Zhi-feng,JI Tong,LI Xiao-wei
DOI: https://doi.org/10.3969/j.1000-6281.2013.02.010
2013-01-01
Abstract:This paper describes a method for automated segmentation of double stained immunohistochemical tissue images that overcomes the limitations of the manual approach as well as of the existing techniques.The method is based on color separation and morphological processing.Compared with the commonly used color deconvolution method,this method improves the segmentation accuracy and only needs to make one section sample instead of three which are necessary for color deconvolution.Experimental results on the tissue images of the squamous cell carcinoma of tongue labeled with cytokeratin(CK) and S-100 protein,stained with 3,3′-Diaminobenzidine(DAB) and 3-Amino-9-Ethylcarbazole(AEC) show that this method can effectively segment cancer cells and nervous cells.In addition,this experiment demonstrates high degree of accuracy and precision compared with manual segmentation.
What problem does this paper attempt to address?