Publication

Abstract : Cervical cancer is the second most prevalent cancer among women in India and has a high fatality rate. Cervical cancer is currently diagnosed via a Pap Smear test followed by a biopsy. The cells from the cervix are obtained and examined for abnormal cells manually in this procedure, which is quite strenuous. Moreover, given the ratio of healthcare professionals to patients, a computer-assisted approach would certainly help pathologists automate the screening process and make predictions much faster. This research paper proposes a system that uses a deep learning model that implements transfer learning with EfficientNet architecture to classify the pap smear cell images. When compared to alternative models for the same accuracy, the EfficientNet architecture has the advantage of using fewer parameters. The work uses the data set “Liquid based-cytology Pap smear data set for automated multi-class diagnosis of pre-cancerous and cervical cancer lesions” that contain images collected from 490 patients from a hospital so as to deal with real-world data. These images are classified into four categories: NILM, LSIL, HSIL, and SCC, with the proposed model achieving a state-of-the-art accuracy of 98.85%.