| Beyza Akyıldız, Caner Özcan Deep Learning Based Cervical Cell Classification from Pap Smear Images |
|---|
| Abstract. Cervical cancer is a significant global health problem that claims many lives each year. Early detection and prompt intervention are critical to improving patient outcomes and reducing mortality rates. This study explores the potential of deep learning models to accurately classify cervical cancer cases using the SIPaKMeD dataset. Several data preprocessing techniques are employed to enhance model performance, with the goal of reducing mortality rates through early and highly accurate diagnosis. The study investigates the effectiveness of transfer learning, hyperparameter optimization, and advanced data augmentation techniques on ResNet101, ResNet50, AlexNet, VGG16, and InceptionV3 models to improve accuracy and generalizability. Compared the other models, these methods were applied to 5 classes in the SIPaKMeD dataset. ResNet_50 achieved the best validation accuracy (0.9547) compared to the other applied models. With this performance output, the proposed method has shown that it is promising in classifying cellular differentiation that can lead to cancer in the cervix compared to other high accuracy approaches in the literature. The results aim to contribute to the existing literature by providing new findings compared to previous methods. The objective is to enhance early detection, improve patient outcomes, and ultimately support the fight against cervical cancer by developing the most effective model. |
| Keywords: Cervical Cancer, SIPaKMeD Dataset, Data Preprocessing, Deep Learning, Classification |
 Download PDF |
| DOI: https://doi.org/10.54381/itta2024.08 |