Analysis of deep learning methods in diabetic retinopathy disease identification based on retinal fundus image

Diabetic retinopathy (DR) is a serious retinal disease and is considered the leading cause of blindness and is strongly associated with people with diabetes. Ophthalmologists use optical coherence tomography (OCT) and retinal fundus imagery to assess the retinal thickness, structure, and also detecting edema, bleeding, and scarring. Deep learning models are used to analyze OCT or fundus images, extract unique features for each stage of DR, then identify images and determine the stage of the disease. Our research using retinal fundus imagery is used to identify diabetic retinopathy disease, among others, using the Convolutional Neural Network (CNN) method. The methodology stage in the study was a green channel, Contrast Limited Adaptive Histogram Equalization (CLAHE), morphological close, and background exclusion. Next, a segmentation process is carried out that aims to generate binary imagery using thresholding techniques. Then the binary image is used as training data conducted epoch as much as 30 times to obtain an optimal training model. After testing, the deep learning method with the CNN algorithm obtained 95.355% accuracy in the identification of diabetic retinopathy disease based on fundus image in the retina.