Overlapping Cervical Nuclei Separation using Watershed Transformation and Elliptical Approach in Pap Smear Images

Abstract

In this study, a robust method is proposed for accurately separating overlapping cell nuclei in cervical microscopic images. This method is based on watershed transformation and an elliptical approach. Since the watershed transformation process of taking the initial seed is done manually, the method was developed to obtain the initial seed automatically. Total initial seeds at this stage represents the number of nuclei that exist in the image of a pap smear, either overlapping or not. Furthermore, a method was developed based on an elliptical approach to obtain the area of each of the nuclei automatically. This method can successfully separate several (more than two) clustered cell nuclei. In addition, the proposed method was evaluated by experts and was proven to have better results than methods from previous studies in terms of accuracy and execution time. The proposed method can determine overlapping and non-overlapping boundaries of nuclei fast and accurately. The proposed method provides better decision-making on areas with overlapping nuclei and can help to improve the accuracy of image analysis and avoid information loss during the process of image segmentation.