Enhanced performance through machine and deep learning methods to quantify PAP smear based cervical cancer detection

Abstract

Cervical cancer is the malignant type tumour among women. It newlineranks as the fourth leading cause of cancer-related mortality. Hence, its newlinedetection and diagnosis at a correct stage is very important to expand the life newlinespan of the women patients around the world. This cancer can be detected newlineusing two methods as Cervigram and Pap smear cell test. Cervigram method newlineis based on detecting the abnormal regions in cervical images. The early newlinedetection is not possible in cervical images. Pap smear cell test is another newlinemethod for detecting the cervical cancer, which detects the number of newlineabnormal nucleolus. newline This research work proposes a methodology for screening the newlinecervical cancer using Pap smear cell images. This method contains newlineenhancement, Gabor transformation along with feature extraction and newlineclassification modules. The features from the Gabor magnitude image are newlinenow optimized using Genetic Algorithm (GA) technique. Finally, the newlineoptimized sets of features are classified using Adaptive Neuro Fuzzy newlineInference System (ANFIS) classifier. The classification rate of this method is newlinenot optimum and hence it has to be enhanced in order to obtain the higher newlineclassification rate. newline Further, the research work develops the automatic detection and newlineclassifications of Pap smear cell images using the modified deep learning newlinearchitecture. The Pap smear cell images are data augmented using shearing newlinefunction. newline