Detection of Cancer in Cervical Cytology Images using Texture Features

This work aims at developing a novel approach for the detection of cervical cancer in Pap smear images using textural feature analysis and provides a direct way to compare those images. The traditional methods employed for this detection would take lot of manual intervention which builds unwanted human errors. The main aim of this paper find out the most relevant texture features by which the detection of cervical cancer can be done, which in turn minimize the manual screening process in detection. This analysis helps in automation process in detection and classification of cytology images based on texture features. It also provides the valuable data to the cytologists to detect the abnormal region in cervical cancer cells. The project is implemented in MATLAB®, a numerical computing environment for data visualization and analysis. Experiments were conducted by extracting various texture features of pap smear images and it is programmed to detect the cancer of normal and abnormal stage in cervical cancer cells. Selection of classifier is also the most important task of this analysis since different classifiers works differently on applying selected features. Applying the most relevant features to various classifiers, we used the Weka tool to rank the features which are mostly fit into the detection. The comparison can be done on classifiers and the optimal results are shown.