Intelligent Diagnosis System For Diabetic Retinopagthy Based On Deep Learning

With the gradual improvement of people's living standards and the increasing life span,the patients with diabetes are also increasing.Diabetic retinopathy(DR)is one of the complications caused by diabetes.Diabetic retinopathy can damage retinal vascular wall,cause retinal exudation,bleeding plaques,capillary hemangioma and other injuries,eventually resulting in decreased vision or even blindness.Diabetic retinopathy with glaucoma,cataract and hypertension are classified as the four major blindness factors.According to the statistics of the office of national health and Family Planning Commission,China is the largest country of type 2 diabetes in the world.The incidence and blindness rate of retinopathy is also increasing.In 2017,the number of diabetic retinopathy patients was 27 million in China.Therefore,it is urgent to treat this disease.On the other hand,research shows that intervention in early stage of diabetic retinopathy can effectively reduce the risk of blindness in patients[1],so early diagnosis of diabetic retinopathy is very important for preventing and treating diabetic retinopathy.At present,the most common method for the diagnosis of diabetic retinopathy is using fundus image.Fundus image is captured by digital fundus camera.It can accurately record fundus information.It is convenient for doctors to observe retinal exudation,bleeding and hemangioma.However,the diagnosis of fundus images requires a rich priori knowledge of the diagnoses,which can be used to determine the condition of the disease accurately.The Chinese population is vast,the level of ophthalmologists is uneven,and the medical resources in remote areas are scarce.The doctors with insufficient experience may fail to diagnose,which makes the patient miss the best time to treat,resulting in serious loss of patients.At the same time,according to the doctor's recommendation,diabetic patients need regular examination of retinopathy.Due to the large number of diabetic patients in China,a lot of fundus images have been generated,which requires the experienced doctors to diagnose,which is time-consuming and laborious.In view of the above problems,the main contents of the work are described below.(1)Artificial detection not only needs experienced doctor,but also is time-consuming and laborious.For solving this problem,this paper uses deep learning techniques to automatically classify lesions.In recent years,with the continuous development of the science of artificial intelligence,the convolutional neural network has made remarkable achievements in the field of image recognition and classification,it not only simplifies the traditional machine learning method,realize the function of end-to-end image classification,but also greatly improve the accuracy.This paper utilizes convolution neural network for classification and diagnosis of diabetic retinopathy,so as to alleviate the problem in a certain extent.(2)In view of the uneven distribution of medical resources,this paper uses the Internet technology to realize remote diagnosis.The traditional remote diagnosis requires patients to upload medical information and fundus image,and then the doctor make online diagnosis.But it still depends on the professional doctor.Because the number of patients with diabetes are numerous,the doctor is exhausting,and lack of medical resources in remote areas can also cause misdiagnosis.In order to solve this problem,this paper is based on B/S architecture,using Nginx + SpringMVC + MyBatis + MySQL + Redis to build a distributed background system,combined with Caffe deep learning platform,constructing a complete set of remote intelligent diagnosis system for diabetic retinopathy.