Segmentation of Pap Smear Images for Cervical Cancer Detection

Abstract

Detection and segmentation of nuclei in pap smear cell images is necessary for automatic screening of cervical cancer. The methodology proposed here is based on K means clustering for segmentation of preprocessed images and is carried out on the Herlev dataset. IoU values of segmentation results with the ground truth are checked and various shape features are extracted from the segmented nucleus. The classification of the nuclei on the basis of the shape features is done with the help of Random Forest Classifier and comparison is made with other classifiers and their results on this dataset.