TransPapCanCervix: An Enhanced Transfer Learning-Based Ensemble Model for Cervical Cancer Classification

Abstract

Cervical cancer, like many other cancers, is most treatable when detected at an early stage. Using classification methods helps find early signs of cancer and small tumors. This allows doctors to act quickly and offer treatments that might cure the cancer. This paper presents a comprehensive approach to the classification of squamous cell carcinoma (SCC) leveraging a dataset comprising 1140 single-cell images sourced from Herlev. In addition to that, in this work, a new ensemble model based on the transfer-learning (TL) technique is developed on various deep learning models, including DenseNet121, DenseNet169, InceptionResNet, XceptionNet, ResNet50, and ResNet101 to demonstrate their efficacy in classifying diverse cellular features. To evaluate our proposed approach's performance, the ensemble approach's results are compared with some transfer learning models such as DenseNet121, DenseNet169, InceptionResNet, XceptionNet, ResNet50, and ResNet101. The experimental results demonstrate that transfer learning-based deep neural networks combined with ensemble methods enhance the diagnostic accuracy of SCC classification systems, achieving 98% accuracy across various cell types. This further validates the effectiveness of the proposed approach. A comprehensive investigation yields a precise and efficient model for SCC classification, offering detailed insights into both normal and abnormal cell types.