Efficient Segmentation of Cervical Cancer Using Deep Learning Techniques

Cervical cancer is a major health concern, and healthcare images play a major role in the analysis and handling of this disease. Three popular deep learning models that can be applied for cervical tumor identification and segmentation. The 3D U-Net model is a customized version of the standard U-Net framework, designed to handle 3D medical imaging data. DeepLab v3+ is another popular semantic segmentation model that uses atrous convolution to confine multi-scale related data. RPN is a popular object recognition model that applies a deep CNN to propose candidate regions in an image that may contain an object of interest. The cervical Cancer Risk Classification Dataset is collected from UCI Repository for assessment of the suggested DL models. The outcome of the DL models is evaluated based on the Dice Similarity Coefficient (DSC), Hausdorff Distances (HD) and Kappa Score (KS). Among the three models 3D U-Net provides better outcomes based on measured output metrics and gave DSC of 0.996, KS of 0.820 and HD of 9.7526 respectively. The tool used for execution is Matlab.