Research On Cervical Cell Segmentation Algorithm Optimization Based On R-CNN

Early screening of cervical lesions is of great significance in pathological diagnosis.Owing to the complexity of cell morphological changes and the limitations of medical images,accurate segmentation of cervical cells is still a challenging task.Computer-aided diagnosis(CAD)is one of the hotspots in the study of cervical cancer screening cell detection.However,there are some problems in the process of cervical cell identification,such as difficulties in accurate segmentation of overlapping cells,blurred cell edges,and poor generalization ability of detection algorithm.In order to solve the problem of complex cervical cell image segmentation,the convolution neural network model is deeply studied,and the image segmentation algorithm is improved and optimized.In the segmentation task,it can be seen from the cervical cell image data set studied that small targets occupy the majority,so we focus on the detection effect of small target when using COCO evaluation index.we focus on some specific issues in cervical cell segmentation.First,based on the Mask R-CNN model,the convoluted underlying features were reused to fuse attention mechanisms and multiscale features to resolve the unclear margins of cervical adhesion cells.The proposed framework was evaluated on the ISBI2014 cervical cell segmentation competition public dataset,and the results show that the average precision(AP)reaches 80.8%,and the small target precision(AP_S)reaches 76.8%.Secondly,combining the advantages of end-to-end instance segmentation,thesis presents an isomorphic multi-branch modulation deformable convolution residual model to solve the inaccurate segmentation of overlapping cell clusters.In addition,instance segmentation is realized based on candidate regions,for obtaining a more accurate prediction box,the regional feature extraction,boundary box recognition,and adding a single pixel-level mask at the last level are integrated and optimized based on the cascade regional convolution neural network,and the functional test is completed,and the segmented effect map of cervical cell instance is obtained.We also use the ISBI2014 dataset to evaluate the optimized model.Experimental results show that the average accuracy of the network model in cervical cell segmentation is81.1%,and the accuracy of small targets is 77%.To some extent,it can assist pathologists in screening cervical cancer in the early phase.