Research And Implementation Of DR Classification Based On Deep Learning

Diabetes mellitus has become the second major chronic disease in China,and its most common complication is retinopathy.Diabetic retinopathy can not be completely cured,the early symptoms of blurred vision,can lead to blindness,so early screening is very necessary.At present,the diagnosis of diabetic retinopathy is mainly through manual screening,which undoubtedly tests the clinical experience of doctors,and inexperienced doctors may lead to misdiagnosis.Therefore,it is of great significance to use computer technology to assist doctors in diagnosing diabetic retinopathy.The main work of this paper is as follows:(1)B-ResNet model is proposed in this paper.In this model,ResNet50 was used to extract the features,CAM was used to strengthen the features,bilinear feature fusion algorithm was used to fuse the feature matrix,and finally the fused feature matrix was classified.To verify the validity of the model,EyePACS,Messidor-2 and IDRiD data sets were used to measure and analyze the experimental results using accuracy,average accuracy,Kappa coefficient,macro-F1 and micro-F1.(2)This paper designs and implements the sugar net auxiliary diagnosis and treatment system based on B-ResNet.Through the analysis of business requirements and user requirements,the functional requirements and non-functional requirements of the system were defined,and on this basis,the system architecture design,functional module design and database design were completed.Spring Boot,Flask and Vue frameworks were used to encode the implementation according to the design scheme.Finally,the system was tested for functional and nonfunctional purposes.The experimental results show that B-ResNet has achieved good results in the experimental data set,and can effectively distinguish fundus images with mild lesions.The sugar network assisted diagnosis and treatment system based on its realization can not only provide doctors with grade diagnosis results of diabetic retinopathy,but also has functions such as identity authentication,personal center,registration appointment and report management,which has certain practical value.