Cervical Cancer Detection Using Advanced Deep Neural Networks

Abstract

One such cancer with high rates of mortality among women in the world is cervical cancer which is mainly caused by late diagnosis and poor access to effective screening procedures. Timely diagnosis is very important as early detection is one of the best ways of improving patient outcome. This paper focuses on application of deep learning techniques in cervical cancer detection of digital cytology images in a fully automated diagnostic system. A CNN model was provided and trained on publicly available datasets of cervical cells and data preprocessing, augmentation, and feature extraction were presented to perform better as a neural network. The standard metrics applied to the proposed system are accuracy, precision, recall, and F1-score, which were used to evaluate the proposed system. As experimental findings reveal, the deep learning model is better when compared to conventional machine learning models and provides high accuracy and strong classification of abnormal and normal cervical cells. The paper also makes a comprehensive study of various network structures with emphasis on how transfer learning can be used to increasing the accuracy of the detection. The study highlights the possibilities of the deep-learning-based methods in clinical practice and adds to the more credible, effective, and scalable solutions in cervical cancer screening.