Cervical Cancer Classification Using Transfer Learning with Hybrid SVM Kernels and ResNet-50

Cervical cancer is a serious health issue and a leading reason of cancer-related deaths among women, particularly in less economically developed countries (LEDCs) where there is limited availability of screening. Early detection is crucial to improving survival rates and quality of life. In this study, we propose an enhanced Support Vector Machine (SVM) model for automated cervical cancer detection, utilizing a novel hybrid kernel that combines the radial basis function (RBF) kernel with a secondary kernel (polynomial). The hybrid kernel approach enables the model to capture both local and global patterns in cell morphology, improving its sensitivity to subtle cell abnormalities and improving diagnostic accuracy. Additionally, we incorporate a pre-trained transfer learning model, Resnet-50 for feature extraction reducing dependency on large labeled datasets while improving robustness. Experimental results demonstrate that the hybrid kernel SVM with Resnet-50 achieves an accuracy of 92%, outperforming traditional SVMs (67%)and hybrid kernel SVMs without Resnet-50 models (69%).