Analysis of features to distinguish epithelial cells and inflammatory cells in Pap smear images

In this work, we propose a novel method for the automated detection of cervical ephitelial cell numbers in Pap smear images, which may contain overlapping nuclei and inflammatory cells. There are three main phases to detect the number of nuclei in this paper. First, the detection of the nuclei areas is based on a morphological image and segmentation of the nuclei boundaries. Second, the shape, the texture and the image intensity are extracted from the nuclei regions and selected with a feature selection scheme based on Feature Subset Selection with Backpropagation classifier for the elimination of false positive findings. At last, Fuzzy C-means clustering algorithm applied on the resulted centroids in order to distinguish the nuclei of cells with inflammatory cells. We evaluated the results by comparing the pathologist rating with respect to the sensitivity and specificity rates. Our proposed methodology is promising with the sensitivity rate of 95% and specificity rate of 98%.