Semi-Automatic Segmentation of Overlapping Cells in Pap Smear Image

Cervical cancer can be prevented only at early stage. Regular pap-smear tests generally recommended by doctors for cervical cancer detection can be more costly for people of developing nations like India. As the number of cancer cases is increasing day by day, so it is a challenging task for conventional medical diagnosis system. In this research paper we have proposed a computer based segmentation approach to segment cells of pap-smear image that can be used for feature extraction and to develop reliable cervical cancer detection system. Segmented cells are used to extract features like nucleus area, cytoplasm area that will give the indication for changes in cell characteristics and can be used in diagnosis system to prevent cancer. Our proposed approach uses a benchmark dataset of pap-smear images and will segment most of overlapping cells from each of the classes.