Classification and Prediction Retinal Oct Images by CNN Algorithm

Abstract

WHO says diabetic retinopathy is a leading cause of blindness. Complexity and quietness make early detection challenging. DR treatment is stage-dependent. Treatment slows disease progression. Ophthalmologists may initially just monitor DR. Diet, blood sugar, and exercise advice must be followed by DR patients. Slows disease progression. Injected medication can reduce DR damage. Advanced RD causes macula bleeding and swelling. Resulting macular edema. Photocoagulation stops retinal leakage. Lasers plug blood leaks by burning abnormal arteries. In the past decade, characterization has improved. We'll use MATLAB's CNN to detect diabetic retinopathy early. CNN is a well-known social model. Input, neurons, stored layers, and output make up CNN's master structure. Both solid and diabetic retinopathy fundus images are well-lit to reveal all hidden details. Mean, standard deviation, variance, skewness, and kurtosis. Bookkeeping is done after extraction. One secret layer, 16 information neurons, and 2 solid or not results make up a convolutional brain network. 70% is used for teaching, 15% for testing. Execution time varied on emphasis or age and location accuracy was 98.93 %. 98.24% exactness, 98.93% accuracy, 99.42% review, and 98.91 % AUC were reported for diabetic retinopathy identification and characterization. © 2022 IEEE.