Three segmentation techniques to predict the dysplasia in cervical cells in the presence of debris

Abstract

Cervical Cancer is one of the most pandemic causes of cancer related death in females. Pap smear test is one of the most commonly used screening test for the cervical cancer. Existing algorithms focus on the segmentation of nucleus and cytoplasm either using single-cell images or multiple cells images. Images captured from the Pap smear slides are called smear images. Smear image contains cervical cells along with debris, debris are inflammatory cells, red blood cells, dye. Debris significantly influence the outcome of image segmentation. An accurate nuclei segmentation method can improve the success rate of cervical cancer screening. Therefore, this paper reveals about three segmentation techniques which are used for automated segmentation of cervical cell nuclei in the presence of debris. Three segmentation techniques namely, Automated Seed Region Growing, Extended Edge Based Detection and Modified Moving k-means techniques are proposed to extract the cervical cell nuclei. These techniques are extracting the area of nuclei from smear images using the morphological property of nucleus. Some debris have area that corresponds with the area of nucleus of normal cells, it may interfere with outcome and may give false positive results. The empirical area threshold value demonstrate the superior performance of all proposed methods. The qualitative and quantitative analysis also done on proposed techniques. Experimental analysis shows that Modified Moving k-means give favorable result in dysplasia detection in the presence of debris. A new dataset PapsmearJP is collected during the study with the help of a pathologist for the validation of the work.