Research And Application Of Histopathological Image Recognition In Rectal Cancer Based On Capsule Network

With the rapid development of medical technology,medical imaging got a massive expansion,scientific use of medical image analysis technology,efficient and accurate classification of tissues and cells image,help doctors better explore ways to treat cancer is medical imaging analysis in recent decades since the most basic one of the most active research areas by classifying tumor tissue on a cellular level can better understand the characteristics of the tumor pathological changes,to help patients better choice method for the treatment of cancer because of cell shape is not limited to strength.It is a challenging task to classify the tissues and cells from the histopathological images of rectal cancer.Purpose of this article is the use of artificial intelligence technique for colorectal cancer pathology image feature extraction and analysis,and then to colorectal cancer pathology image recognition and classification at present,the use of artificial intelligence technology to the pathology image classification research mainly concentrated in two aspects,on the one hand is a man-made feature extraction combined with traditional machine learning algorithms of pathology image classification,another aspect is based on the deep learning pathology image classification based on the depth study of image classification can automatically learn from annotated image data set to the complex.The characteristics of the higher level,to avoid the manual designed to extract the characteristics of filters,limitations and complexity but currently there exist certain disadvantages to the convolution of deep learning neural network,to improve its shortcomings,this article will use the capsule network technology forecast and diagnosis of colorectal cancer tissue pathology image,this paper innovation points are as follows:(1)In view of the convolutional neural network training process issue of space structure levels,this paper puts forward an improved capsule network architecture,using VGG as feature extraction,the characteristics of the constructed graph has a largerreceptive field,can obtain more abstract capsule of high-level features and then through the capsule network layer code to the extraction of the high-level features,combined with dynamic routing algorithm with the digital capsule layer,the relationship between the network to enhance capsule The experimental results show that VGG can effectively extract the key information of pathological image features,and the dynamic routing of capsule network can make features more spatial and directional.(2)Aiming at the problem of medical image with high cost,this paper proposes a model to combat model was generated based on Caps Net capsule network as implicit vector auto-encoder,each input kinds of real samples correspond to different vector space,noise will be through the generator is mapped to the implicit vector space,then through data generated to resume samples,followed by using discriminant to authenticity discriminant samples and two model of game,finally realizes the target cost function minimum to verify the validity of the model,the paper successively in histopathological image data sets for rectal cancer were trained and compared with other experiments.(3)To promote the development of machine learning in the field of medicine,medical aided diagnosis system is designed for the diagnosis of the disease in system development,combined with pathology experts experience and colorectal cancer TNM staging system disease as computer quantitative standard for the disease,through pretreatment module texture feature description module image texture feature analysis module and pathology diagnosis module,diagnosis results generally conclusive summary,summarized characteristics of disease as the judgment basis was provided for the preliminary diagnosis of the disease.To some extent,the system has solved the problem of delayed diagnosis and treatment in some regions where medical treatment is scarce,and has solved the problem of misdiagnosis caused by the influence of experts’ clinical experience and personal subjective factors,thus saving the precious time for the treatment of patients.