A Novel Method for Segmenting and Ranking Nuclei for Cervical Cell

摘要

The main objective of this paper is to automate segmentation and analyze nucleolus to cytoplasm ratio of cervical cell. The first step, the nuclei initialization, concentrates on the target nuclei and related boundaries by removing incidental regional minima from the gradient map. The multi-scale gradient watershed transform is used for nuclei segmentation. It identifies regional minima and preserves the valid nuclei. A local filter is designed for the segmentation of cytoplasm. The performance of the proposed method measuring Nuclear to Cytoplasm (NC) shows better results as compared to other any other methods available in the literature.