Fuzzy Cells Extraction and Cell Morphology for the Antigen Positivity Determination of Immunohistochemical Image

The accuracy of immunohistochemical image antigen determination is very important in reporting result immunohistochemical examination. The previous studies [5], [10] have not been succeeded to determine such accuracy. The failure is caused by the characters in the image cells vary widely in shape, size, cell type, outward color appearance, and the huddled cells. EsFuMos method that uses fuzzy inference system is proposed to extract the cells, obtain cell morphology, and watershed separating huddled cells. The experimental result of comparing EsFuMoS method with MoFSoTH method [5] and the NN method [10] shows that EsFuMoS method is the most successful method to determine positivity. Meanwhile for huddled cell separation, EsFuMoS method success rate is 77.63%. Lack of EsFuMoS method is having an error in cell separation, and not able to lift the weak negative cell nucleus. According to pathologist, this method works very helpful. But it needs to be improved by paying close attention to the character of the cell membrane and weak cell nucleus.