Medical Image Registration at Pap Smear for Early Identification of Cervical Cancer

Abstract

The complexity of the cell structure and high overlap causes poor image contrast. Complex imaging factors in lighting differences, dye concentrations, and other variables such as drying air, excess blood, mucus, bacteria, or inflammation can make automatic visual interpretation more difficult. This study proposes an approach model by combining basic image processing techniques in deep learning for segmentation of the nucleus in the Overlap Cell Image of Pap Smear of Cervical Cancer patients. The purpose of this research is to segment by increasing the identification accuracy of Pap smear images on RepomedUNM public data. The results have the best performance as seen in the MSE value, the lowest RMSE value is 0.2024253 and the lowest PSNR is 0.04009707 and the highest PSNR is 65.3826018 dB. So, this study can be used as a reference in identifying the Cervical Cancer Nucleus as Medical Image Registration (MIR) patients.