Research On Intelligent Recognition Method Of Breast Cancer Pathological Images

Histopathological images are the "gold standard" for cancer diagnosis.The resolution of full-slice histopathological images is very high.Pathologists need to distinguish pathological images based on their own accumulated experience and professional knowledge.They are subjective and have low diagnostic efficiency and cannot be repeated.Computer-aided diagnosis can effectively avoid the subjectivity of diagnosis and improve the efficiency and accuracy of diagnosis.In this paper,the pathological image of breast cancer is the research object,the computer-aided classification and segmentation methods of pathological images are carried out,and the intelligent diagnosis platform of breast cancer pathological images is built.The main research contents of the thesis are as follows:(1)The classification of pathological images of breast cancer is realized based on deep learning.In order to solve the problem that the classification network cannot train the WSI,a fixed-size image block is extracted to obtain the training samples and labels of the classification network,and the model is trained.Model performance largely depends on high-quality training samples.For this reason,a combination of Ostu algorithm,high-pass filtering and small connected region elimination algorithm is proposed to preprocess the image,and segment the background and tissue area of the pathological image,which is very important to obtain the image block with high discriminative power.At the same time,the extracted image blocks are stained and standardized.Two popular network models are used for comparison.In terms of efficiency,the Mobile Net V2 model has great advantages;in terms of accuracy,the Res Net101 model has higher classification accuracy.To this end,you can select the corresponding network model according to the needs of different scenarios.(2)Established a segmentation model of breast cancer pathological images.On the basis of the classification network with good performance,the segmentation processing of histopathological images is carried out.According to the idea of ensemble learning,a two-level network model is proposed.The classification model based on the image block level is first used for fast and rough classification,and the confidence interval is reduced.Then,the pixel-based semantic segmentation model is used for precise segmentation.The features extracted by the sub-network realize the precise segmentation of the lesion area of the pathological image of breast cancer,and further improve the performance of the convolutional neural network in the location of the lesion in the pathological image of breast cancer.(3)In order to further promote the intelligentization and digitization of breast cancer pathological images,this paper developed an intelligent recognition platform for breast cancer pathological images based on the Web-based B/S model.The main functions of the platform include image query module,intelligent diagnosis module and image management module.Through the intelligent diagnosis module,users can upload images,perform intelligent diagnosis of breast cancer pathological images,quickly obtain diagnosis results,and provide users with diagnostic references.The establishment of an intelligent recognition platform for pathological images of breast cancer stores the diagnosed pathological images of breast cancer in the database.It is an innovative medical service software for medical image assisted diagnosis and medical image management,which assists doctors in pathological image diagnosis.The establishment of a pathological image resource library,a system that facilitates centralized image management,can speed up pathological image diagnosis,reduce the workload of pathologists,and realize intelligent management of pathological image data.Through the research on the intelligent recognition method of the pathological image of breast cancer and the discussion on the construction of the intelligent diagnosis platform,it provides guidance for the realization of the digital and intelligent diagnosis of the pathological image of breast cancer.