Classified Study Of Diabetic Retina Based On Deep Learning

At present,China is the country with the largest number of diabetic patients in the world.As the age of diabetes increases,many complications will occur.Retinal disease is one of the most threatening congenies in diabetic patients,which seriously affects the quality of life of diabetic patients.Since the classification of the degree of retinal disease lesions is a complex process,and the classification definition is relatively vague.How to accurately diagnose the degree of retinal disease lesions is a major problem faced by ophthalmologists.Another major problem is the apparent imbalance between ophthalmologists and patients with diabetic retinopathy,especially in remote township areas.Therefore,this paper studies the classification of diabetic retina based on deep learning algorithm.The specific research contents include:The scheme used for the inaccurate classification of the degree of retinal disease lesions: First,a pre-processing step is performed for defects existing in a particular color fundus image,including color balance adjustment,black background removal,cropping to a uniform pixel value,and flip image expansion data set.Then a deep learning correlation algorithm is used to apply a two-class model based on the residual network and another five-class model to improve the Jeffrey scheme,and the network model is trained by using the pre-processed data set.The simulation results verify that the trained model can be used to diagnose the degree of retinal disease lesions,and the classification results can be used as a reference for the ophthalmologist to diagnose the degree of the lesion.For the imbalance between ophthalmologists and diabetic retinopathy patients in remote areas: Considering that the number of ophthalmologists is unlikely to increase immediately,patients with retinal diseases are increasing year by year.Therefore,this thesis designs a diagnosis page for diabetic retinal diseases by building a BOA server and using HTML technology,and then uses CGI script program design to diagnose whether the color fundus image is sick and can give the result of grading prediction.Applying the diagnosis webpage to township clinics in remote areas can solve the problem of serious imbalance between doctors and patients to a certain extent.For such a large diabetic retinal patient in China,if the degree of disease cannot be detected and determined in time,there is a possibility of blindness.However,the lack of clinical experience of ophthalmologists who have just entered the post and the lack of medical resources in remote areas will delay the patient's condition and even lead to misdiagnosis.In addition to assisting doctors in making diagnostic results,the research results of this thesis can also solve the current situation of lack of ophthalmologists in remote areas.