DeepCerviCancer-Deep Learning-Based Cervical Image Classification using Colposcopy and Cytology Images

Abstract

INTRODUCTION: Cervical cancer is a deadly malignancy in the cervix, affecting billions of women annually. OBJECTIVES: To develop deep learning-based system for effective cervical cancer detection by combining colposcopy and cytology screening. METHODS: It employs DeepColpo for colposcopy and DeepCyto+ for cytology images. The models are trained on multiple datasets, including the self-collected cervical cancer dataset named Malhari, IARC Visual Inspection with Acetic Acid (VIA) Image Bank, IARC Colposcopy Image Bank, and Liquid-based Cytology Pap smear dataset. The ensemble model combines DeepColpo and DeepCyto+, using machine learning algorithms. RESULTS: The ensemble model achieves perfect recall, accuracy, F1 score, and precision on colposcopy and cytology images from the same patients. CONCLUSION: By combining modalities for cervical cancer screening and conducting tests on colposcopy and cytology images from the same patients, the novel approach achieved flawless results.