On the importance of nucleus features in the classification of cervical cells in Pap smear images

In this work, we investigate the classification of cervical cells by exploiting only the nucleus features and not taking into account the features extracted from the cytoplasm area. This procedure is motivated by the fact that the nuclei areas can be extracted automatically from Pap smear images, in contrast to the cytoplasm segmentation which is not a solved problem yet. Furthermore, we consider the representation of these features in low dimensional spaces using non linear dimensionality reduction techniques, which can properly handle complex nonlinear data, as they better describe their manifold structure. The classification was performed in a supervised manner, using support vector machines (SVM) with several kernel functions. The obtained results indicate that we can achieve high classification performance when we use features extracted only from the nuclei areas.