ABSTRACT
Objective To investigate an effective algorithm for image segmentation in cervical cancer cell adhesion , which enables accurate segmentation of the contour of adherent cells .Methods The images of target cells were extracted from the background area using level set methods , normalized with minimum values of transformation algorithms ,and multi-plied by the gradient image points in the region of interest ( ROI) to inhibit the undesired gradient information before the im-ages of adherent cells were segmented using labeled watershed algorithm .Results and Conclusion Compared to conven-tional watershed segmentation methods , this algorithm is not only effective in image segmentation of adherent cervical cancer cells with uneven staining and more accurate segmentation lines established around the contours of adherent cells , but of high clinical value .