Enhancing Cervical Cancer Diagnosis using Deep Learning based CAD Systems

Cervical cancer remains one of the most common and life-threatening diseases in women worldwide. Early and accurate detection is important to improve patient survival rates. In this study, we propose CervicalNet, a deep learning model designed for the automatic classification of cervical cell images to provide real-time assistance in computer-aided diagnosis (CAD) systems. The model is trained and analysed on Mendeley Liquid-Based Cytology (LBC) Image set, a diverse dataset collection of labeled images of normal and abnormal cervical cells. CervicalNet is constructed using a custom designed convolutional neural network (CNN) architecture for the purpose of biomedical image classification. It combines preprocessing methods such as normalization, augmentation, and class balancing to improve generalization and stability. The proposed system was trained and tested on an 80-20 data split, yields classification performance with accuracy of 97%, precision of 96%. This research proves the feasibility and efficiency of using a specialized deep learning model for cervical cancer detection, providing a scalable solution for real-time diagnostic assistance.