A Novel Web Framework for Cervical Cancer Detection System A Machine Learning Breakthrough

Abstract

Cervical cancer remains one of the leading causes of cancer-related deaths in women globally, with early detection being crucial for improving survival rates. However, current diagnostic methods like Pap smears are often subjective, prone to human error, and inefficient in handling class imbalance in medical image datasets. This study aims to develop an improved cervical cancer detection model by combining Convolutional Neural Networks (CNNs) with Support Vector Machines (SVMs) to enhance accuracy, efficiency, and interpretability. The study proposed a hybrid CNN-SVM model to address class imbalance and computational inefficiency. The CNN extracts feature from Pap smear images, while the SVM classifies the images into cancerous or non-cancerous categories. The dataset used includes labeled cervical cell images, and the model’s performance is evaluated using cross-validation and standard metrics such as accuracy, AUC-ROC, F1 Score, and AUC-PR. Additionally, Grad-CAM is integrated to provide model interpretability. The proposed model achieves an accuracy of 95.2%, an AUC-ROC of 0.981, and an F1 Score of 0.93, outperforming existing models, including traditional SVM and CNN-only approaches. The hybrid model demonstrates significant improvements in handling class imbalance and computational efficiency. This research contributes a novel hybrid model that enhances cervical cancer detection by combining deep learning and traditional machine learning methods. The model’s high performance and interpretability make it a promising tool for real-world clinical applications, particularly in resource-constrained settings, improving early diagnosis and patient outcomes.