Research Progress of Diabetic Disease Prediction Model in Deep Learning

Abstract

Diabetes is a metabolic disease characterized by high blood sugar, which is mostly caused by the defect of insulin secretion or the impairment of its biological function. The persistent high blood sugar in diabetes will damage various tissues, especially the brain, kidney, heart, nerves and so on. Diabetic disease is one of the main causes of visual impairment in diabetic patients. Early classification diagnosis is of great significance for the treatment and control of the disease. Deep learning methods can automatically extract the characteristics of retinopathy and classify them, so they become an important tool for the classification of diabetic retinopathy[1-3]. Firstly, the application of deep learning in the binary classification of diabetic retinopathy was summarized by introducing the commonly used data sets and evaluation indicators of diabetic retinopathy. Secondly, the application of different classical deep learning models in the classification of the severity of diabetic retinopathy was reviewed, with emphasis on the classification and diagnosis methods of convolutional neural network, and a comprehensive comparative analysis of different methods was made. Finally, the challenges facing this field are discussed and the future development direction is prospected.