Deep Learning-Based Classification of Cervical Cancer using pap smear images

Cervical cancer, originating in the cervix, poses a significant health concern, ranking as the fourth most diagnosed cancer and leading cause of cancer-related deaths in women. It often remains asymptomatic in early stages, making regular screenings crucial for early detection. Current diagnostic methods involve Pap tests and HPV tests which has challenges in diagnosis which include low screening rates, Pap smear sensitivity, and variability in interpretation. To address challenges, an AI based approach has been done in this paper. The study employs various deep learning architectures, namely ResNet18, ResNet50, GoogLeNet, and SqueezeNet, while carefully considering different epoch settings. The results showease ResNet18 as the top-performing model, attaining the highest test accuracy of 98.51%. The findings emphasize the importance of selecting the appropriate network architecture and training duration, tailored to the characteristics of cervical cancer classification. Future developments in these areas could revolutionize cervical cancer diagnosis, making it more effective and widely accessible.