Comparative Analysis of Deep Learning Pre- Trained Models and Machine Learning Classifiers for Accurate Cervical Cancer Diagnosis using Colposcopy Images

Cervical cancer is a significant public health issue worldwide, and early detection is essential for successful treatment. Colposcopy is a widely used diagnostic tool for cervical cancer, but its accuracy depends on the examiner's experience. Computer-aided diagnosis (CAD) systems can potentially improve diagnostic accuracy and reduce variability among examiners. The proposed system includes feature extraction using different deep-learning pre-trained models and classification into cancerous and non-cancerous images using different machine-learning classifiers. The efficacy of the suggested system is assessed utilizing a conventional dataset comprising colposcopy images. The results of study demonstrate a potential of CAD systems for improving accuracy of cervical cancer diagnosis and reducing inter-observer variability using deep learning. The paper presents that the Googlenet and ResNet-50 pre-trained models features from transformation zone extracted colposcopy images have the highest classification accuracy of 76.64%.