Detection of Cervical Cancer at an Early Stage Using Hybrid Segmentation Techniques from PAP Smear Images

Abstract

© 2018 SERSC Australia. Cervical cancer is the abnormal cell growth on cervix. The lower part of the uterus or womb that opens top of the birth canal is called cervix. Unlike other cancer cervical cancer has no telltale symptoms so cannot be detected at an early stage thus mortality rate is very among woman. PAP smear test is an efficient and popular screening procedure to detect the presence of abnormal cells in cervix. Manual screening is not always perfect, quantifying and analyzing the abnormal cells from the large number of PAP smear images is a tedious task. A novel approach has been proposed to detect and analyze the abnormal cells from PAP smear images at an early stage, that it will help for fast and proper treatment. The proposed method carried out four stages: preprocessing, pre-segmentaion, segmentation and feature extraction. In preprocessing stage, unwanted noise and artifacts have been removed. In pre-segmentation stage, histogram has been computed first then morphological operation followed by edge detection techniques have been performed. In segmentation phase nucleus has been segmented by hybrid segmentation method which consists of thresholding and watershed segmentation. Two types of feature extractions like (a) morphological feature extraction, (b) texture feature extraction GLCM (Gray Level Co-Occurance Matrix) have been performed in this stage for accurate detection of abnormalities in cervical region. This paper presents a complete qualitative assessment which is used to develop a self-constructing algorithm for diagnosis the cancerous cell from PAP smear images.