Median Filter and U-Net Architecture for Robust Segmentation Nucleus and Cytoplasm on Pap Smear

Cervical cancer is a condition caused by a layer of malignant cells that grows and develops rapidly on the cervix due to infection with the human papillomavirus virus (HPV). Cancer detection of the cervix can be done by a pap smear examination. This study aims to build an automatic segmentation model by combining augmentation and segmentation. The augmentation techniques used in this study are flip and median filters. Augmentation techniques aim to increase the quantity and variation of data to improve the quality of segmentation results. U-Net architecture is often used for image segmentation. By combining data augmentation and U-NET architecture, it is expected to meet the needs of this model which requires a lot of data. A combination of augmentation with U-NET is anticipated to improve the performance of the model significantly. The parameters used to measure the performance of the proposed method include accuracy, precision, recall, and Intersection over Union (IoU). The results of this method show an accuracy of 92%, precision of 91%, recall of 90%, and IoU of 84%. The results of the performance evaluation show that the method proposed for pap smear segmentation has excellent and powerful capabilities. However, the IoU performance in this study needs to be improved for further study. The proposed method can be used as a model for the development of automatic cervical cancer detection applications in the medical field.