Automated Cervical Cancer Detection Using Image Processing … Deep Learning Model
Devi. S, Vijayakumar Adaickalam · 2025
Cervical cancer is an important women's health problem worldwide, with a yearly incidence of over 600,000 new cases. Improvements in patient outcomes and decrease in mortality rates is mainly due to early and accurate detection. Pap smears and HPV tests are the standard screening tests. Though effective, these tests heavily rely on expert manual evaluations, which are time-consuming and subject to human error. To overcome these issues, this paper suggests a deep learning system for cervical cell automatic classification based on Convolutional Neural Networks (CNNs). The system is trained on a dataset of 4049 cervical cell images, labeled into five common classes: Dyskeratotic ($\mathbf{8 1 3}$images), Koilocytotic ($\mathbf{8 2 5}$images), Metaplastic (793 images), Parabasal (787 images), and Superficial-Intermediate cells (831 images). The aim of this system is to assist pathologists with an effective, consistent, and accurate diagnostic tool, thereby reducing their workload and improving clinical decision-making processes. These experimental results show that the model attains a mean classification accuracy of$\mathbf{9 6. 0 3 \%}$, indicating strong potential real-world applicability in the clinical scenario. This paper highlights the revolutionary potential of artificial intelligence in medical diagnostics and foresees the necessity for improvement in deep learning-based systems for cervical cancer screening.