Iterated Shape-Bias Graph Cut-Based Segmentation for Detecting Cervical Cancer from Pap Smear Cells

Abstract

Cervical cancer is the most common type of cancer affecting the Indian women population. It is determined to be prevented by periodic screening which identifies and treats precancerous lesions in contrast to other cancers. Diversified screening methodologies that include traditional and advanced technologies are available for screening women cervix for the purpose of detecting cervical cancers and pre-cancers. No single screening test is perfect due to the manual error inherent in each cervical cancer detection technique. Computer-assisted approaches that use graph cut-based segmentation is potential in detecting abnormalities from the pap smear cell. In this paper, Iterated Shape-Bias Graph Cut-based Segmentation (ISBGCS) method is proposed for achieving excellence in diagnosing cervical cancer, since they are capable of accurately deriving the boundaries of nuclei and cytoplasm even when the boundaries are hazy and overlapping. This ISBGCS method facilitates effective detection of nuclei and cytoplasm boundaries even under irregular staining of pap smear cell screening process. In specific, it considers local edge consistencies and global shape with shape priors simultaneously to evaluate the segmentation process. The resultant cues are formulated into a graph cut with local and global shape priors for achieving optimal segmentation during cervical cancer diagnosis.