Cervical Cancer Classification From Pap Smear Using XCeption Model

Cervical cancer is a disease that affects the reproductive organs of women and contributes to a high mortality rate among them. The rise in mortality can be attributed to delays in cancer screening, underscoring the importance of early prevention measures. While the Pap smear test is an effective method for early detection and diagnosis of cervical cancer, manual inspection of test results still has limitations in terms of observation and time-consuming laboratory processes. This study aims to implement the Convolutional Neural Network (CNN) model using transfer learning with the Xception architecture to classify cervical cancer based on Pap smear images, categorizing them into 2 or 4 classes. The dataset used in this study consists of 400 cervical cell images, including normal cells, koilocytes, L-Sil, and H-Sil, sourced from RepoMedUNM. The study involves several steps: data splitting, data resizing and augmentation, hyperparameter tuning, CNN model development using the Xception architecture, training, testing, and model evaluation. The research results indicate that the Convolutional Neural Network (CNN) with transfer learning using the Xception method can accurately classify cervical cancer based on Pap smear images, achieving an accuracy of 100% in the 2-class classification and 85% in the 4-class classification.