Diagnosis of pre-cancerous cells in the cervix using statistical and morphological features from Pap smear images

Introduction: Cervical cancer is the second most common disease in women due to abnormal cell growth. The Pap smear test is one of the common test for screening cervical cancer. The aim of this study is automatic detection of the pre-cancerous cells in microscopic images of Pap smear samples using image processing techniques. Research method: In order to gather the collection of images, 61 samples of Pap smear were tried with a Nikon V1 camera connected to the Nikon Eclipse i50 Nikon microscope. The resulted sample space is 1073 microscopic digital images. To differentiate the cells through different colored bands and to increment of the contrast, the pop-smear image split automatically and thresholding by the use Otsu method. In order to featuring, different statistical textural properties are used, such as GLCM and geometric features such as area, extreme diameters, centrifugal, environment and density. In order to differentiate between healthy and malignant cells in this study, a backup vector machine (SVM) has been used. Results: The accuracy of designed algorithm was 94.9% and its AUC was 96.51%. Conclusion: we propose to enhance and develop this Algorithm to aid pathologists in diagnosis of such type of malignancy.