Improving Cervical Cancer Detection Through Machine and Deep Learning Algorithms

Cervical cancer, a prominent contributor to mortality in low-income countries, can often be effectively treated if detected early through the removal of affected tissues. Automating the identification of pre-cancerous cervical lesions using image-based methods represents a critical advancement, revolutionizing the detection of cervical pre-cancerous features. In recent years, the interest of researchers in the fields of computer vision and machine learning has been drawn to the development of computer-aided diagnostic systems for processing digital colposcopies. This has led to a multitude of challenges and computational solutions. In this study, we employ various algorithms, including Convolutional Neural Networks (CNN), Faster R-CNN, Region-based Convolutional Neural Networks (R-CNN), Random Forest, and Extreme Gradient Boosting (XGBoost), Support Vector Machines (SVM), for automated colposcopy-based detection of pre-cancerous lesions. Additionally, we contribute to ongoing efforts aimed at improving early detection and treatment of cervical cancer, particularly in resource-limited settings, by utilizing a mathematical model-based classification method to extract features used in identifying pre-cancerous characteristics.