MobileNetV2 Based Cervical Cancer Classification Using Pap Smear Images

With over 500,000 new cases reported globally each year, cervical cancer stands out as a particularly serious form of cancer, predominantly affecting women. Visual inspection following acetic acid application (VIA), Papanicolaou test (Pap) and human papillomavirus (HPV) test are some of the popular methods used for cervical cancer screening. The study on the Convolutional Neural Network (CNN) architecture uses MobileNetV2 as a transfer learning tool. This lightweight CNN is implemented to automatically classify cervical cancer from medical photos with significant effectiveness and adaptability. This is perceived through the model that it performs well in terms of early classification indicating that it may be included into medical science to improve access to enhance the availability of cervical cancer screening in resource-constrained environments. The best discriminant characteristics for five classes of Pap pictures (cervical cell samples) are obtained through the method. Images from Pap Smears tests offer comprehensive details regarding the morphology—size, shape, and structure of cervical cells which further aids in identifying cellular abnormalities indicative of precancerous or cancerous conditions. The classification accuracy for MobileNetV2 was 95%. Comparing MobileNetV2 to other contemporary neural architectures, it achieved greater validation accuracy while reducing computational cost, thus outperforming existing algorithms in the field of cervical cancer diagnosis. This can assist continued efforts to increase early detection of cervical cancer and eventually enhance the health and well-being of women.