A Comprehensive Review and Implementation of a Novel Hybrid Model for Cervical Cancer Diagnosis
Sandhya Vani Meruva, Pratyusha Desireddy, Devi Sravani, M. Mohan · 2024
Cervical cancer is a prominent health issue and a primary contributor to female mortality from cancer worldwide. Prompt identification and precise assessment are essential for enhancing patient results. Conventional screening techniques, like the Pap smear test, are efficient but suffer from time-consuming procedures and susceptibility to human mistakes. This study seeks to overcome these constraints by utilizing novel learning methodologies to create an automated and dependable diagnostic tool for the detection of cervical cancer. This paper has two main objectives: to offer a thorough examination of current methods used for detecting cervical cancer, and to present initial findings from the application of a new hybrid deep learning model. The proposed model, ReNab-Net, combines ResNet, a widely recognized Convolutional Neural Network (CNN) architecture, with Naive Bayes, a probabilistic classifier renowned for its simplicity and efficiency. This hybrid approach is specifically designed to improve the accuracy of diagnoses. The research methodology encompasses a comprehensive preprocessing pipeline, commencing with the acquisition of datasets containing cervical cancer images. The images undergo bilinear interpolation for resizing, conversion to grayscale, and enhancement through Multi-Peak Histogram Equalization (MPHE) to enhance contrast. The improved images are subjected to gray level thresholding using the Triangle method and edge detection using fuzzy logic methods. Feature extraction is conducted by utilizing Local Tetra Patterns (LTrPs), which capture spatial relationships and directional intensity variations in order to create strong and reliable feature vectors. The ReNab-Net model is developed and framed in the work. This study enhances the continuous endeavors to enhance cervical cancer diagnosis by means of technological advancement, aiding healthcare professionals in delivering prompt and precise treatment. The survey and primary findings provide valuable insights and a strong basis for future research in the field of automated cervical cancer detection.