Automated Semantic Segmentation of Cervical Cells Using High-Resolution Network

With an emphasis on the Cx22 dataset, this work offers a novel method of semantic segmentation applied to cervical cytology images. To precisely segment cervical cell structures within the images, the study makes use of deep learning techniques. In this work we are proposing different deep network models, U-Net, LinkNet, PSPNet and HRNet to perform semantic segmentation of the cells in the cervical cytology images. The performance evaluation of all the models are done using the high quality Cx22 dataset with 14,946 finely annotated cellular instances and the HRNet deep model gives better segmentation results and shows its effectiveness in accurately drawing cell boundaries. The outcomes demonstrate how our method can improve automated cervical cytology analysis with a very good F-score of 92% which leads to improvements in diagnostic support and medical image processing.