Segmentation of immunohistochemical image of lung neuroendocrine tumor based on double layer watershed

Lung neuroendocrine tumor is a special kind of lung cancer. The level of tumor is a major factor in the diagnostic process of this kind of tumor, where Ki-67 proliferation is often used as the classification index. The segmentation and classification of Ki-67 immunohistochemical images of lung neuroendocrine tumors is the key to realize automatic calculation of Ki-67 proliferation index. In this paper, a new image segmentation algorithm based on double layer watershed is proposed, which is suitable for the characteristics of uneven color of the nucleus. For calculating the Ki-67 proliferation index, classification of negative and positive tumor cells by their color difference is accomplished. By means of the features of regional average gray level, area and circularity, the segmentation of nucleus which cannot be completely segmented is supplemented. The mean and standard deviation are used to evaluate the accuracy after segmentation, and they have major improvement through the new algorithm proposed in this paper. Detailed theoretical analysis and reasonable experimental results demonstrate the feasibility and accuracy of the method.