A Colour Space Based Detection for Cervical Cancer Using Fuzzy C-Means Clustering

Abstract

This research presents a colour segmentation method using Hue, Saturation, Value (HSV) colour space based on fuzzy c-means clustering (FCM) to segment nucleus from single cell Pap smear images. Nucleus is a structural part of cell which can indicate whether a cell is normal or abnormal. This research aims to analyze the performance of colour space in the segmentation process. Pap smear images were segmented in HSV colour space by using fuzzy c-means clustering technique. Compared with segmentation process directly on HSV channel, the segmentation of each channel in space H, S and V were proposed. The segmentation results on each channel that has been applied roundness detection subsequently merged as the final segmentation and labeled as a nucleus. This research used 70 single cell Pap smear images taken in harlev dataset to examine the proposed segmentation method. The calculation of segmentation performance used the measurement based on precision, recall, and Zijdenbox Similarity Index (ZSI). The result showed that the proposed method generated precision, recall, and ZSI by 93%, 94%, and 93%.