Manuscript received November 28, 2025; revised January 7, 2026; accepted March 5, 2026; published July 17, 2026.

Abstract—Pap-smear cytology is essential in the detection of cervical cancer at an early stage, but manual cervical cancer examination is time-consuming, subjective, and liable to diagnostic variability. Deep learning provides an alternative (automatic) method based on cytology. The paper introduces an explainable deep-learning architecture to classify cervical cells and the comparison of nine popular Convolutional Neural Networks (CNN) models, namely DenseNet, ResNet, Visual Geometry Group (VGG), Inception, EfficientNet, and a custom CNN-18 model. Our collection contained 6374 pictures of five epithelial classes, used standardized preprocessing and massive augmentation, and trained all the models using the same pipeline. The models were tested based on accuracy, precision, recall, F1-Score, and Area Under Curve (AUC). The highest performance was obtained by DenseNet-121, which has an accuracy of 98.85% and an F1-Score of 0.9885. It was much better than more complicated models, which were proven using ANOVA (p < 0.05). Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations indicated the model is always interested in clinically pertinent nuclear and cytoplasmic regions and, therefore, enhances transparency and facilitates expert interpretability. These results indicate that DenseNet-121 offers an accurate, computationally efficient, and understandable solution that can be used in cervical cancer screening, which is supported by Artificial Intelligence (AI). The framework could be implemented in digital pathology workflows and applied in various clinical settings.
 
Keywords—cervical cancer classification, pap-smear cytology, deep learning, convolutional neural networks, DenseNet-121, transfer learning, computer-aided diagnosis, explainable artificial intelligence, Gradient-weighted Class Activation Mapping (Grad-CAM), medical image analysis

Cite: Sadek Al Prince, Gazi Mehraj Sakib, Mahe Zabin, Shakila Rahman, and Jia Uddin, "Explainable Deep Learning for Cervical Cancer Cell Classification: A Comparative Analysis of CNN Architectures with DenseNet-121," Journal of Image and Graphics, Vol. 14, No. 4, pp. 628-644, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).