Fuzzy-Based Classification for Cervical Dysplasia Using Smear Images

Abstract

Cervical dysplasia is the second most cause of the death in the females. A Pap smear test is the most efficient and prominent screening method for the detection of dysplasia in cervical cells. Pap smear is time-consuming, and sometimes, it is an erroneous method. Automated and semi-automated systems can be used for cervical cancer diagnosis and treatment. In our proposed approach, we are segmenting image first; the RGB image transformed into L * a * b * format. Then, using K-means clustering technique image has segmented into background and cytoplasm. Thresholding and the morphological operations have used to segment nucleus only from the second cluster. The shape-based features of the nucleus have been extracted. In the classification phase, fuzzy C-mean (FCM) has been used for clustering. Principle component analysis (PCA) is used to find the most prominent features. The classification of Pap smear images is based on the Bethesda system. The approach has performed on a dataset obtained from pathologic laboratory containing 150 Pap smear images. Performance evaluation has been done using Rand index (RI). The RI of fuzzy C-mean is 0.933, and using PCA, it is 0.95.