Deep Learning Based Framework for Cervical Cancer Diagnosis through Feature Extraction and Classification

Cervical-cancer continues to be a significant contributor to cancer-related fatalities among women worldwide, highlighting the importance of early identification in lowering mortality rates. Traditional methods of cervical cancer detection, such as Pap smear tests, are highly dependent on the expertise of the examiner, often leading to misclassifications. The problem this work addresses is the need for an accurate, automated system that can classify Pap smear images into multiple classes for cervical cancer detection. Hence, this work proposed a deep learning-based approach using Convolutional-Neural-Network VisualGeometry-Group-16 (CNN-VGG16) for cervical-cancer classification in Pap-Smear images. The methodology comprised preprocessing input images through resizing and normalization, followed by feature extraction using convolutional-layers. The features extracted were then classified utilizing pre-trained VGG16 model. The CNN-VGG16 approach achieved better results, with 99.59% accuracy for 2-class classification, 99.35% for 5-class classification on SIPaKMeD and 99.59% accuracy for 2-class classification and 99.23% for 7-class classification on Herlev datasets. The novelty of work lies in integration of feature extraction with CNN-VGG16, which enhances the model’s performance. The findings also show that CNN-VGG16 method outperforms existing approaches