Diabetic Retinopathy Image Grading Using Deep Learning Based Pipeline

Abstract

Diabetic retinopathy is considered one of the most dangerous diseases that affect the eye and may lead to complete blindness if early diagnosis is not performed and the required treatments are not performed.

It can be recognized by retinal lesions, which are microvascular aneurysms, hemorrhages, and secretions.

Previous image processing methods for diabetic retinopathy have been able to manually diagnose the degree of the lesion, but are timeconsuming and inaccurate.

In our study of this topic, a research framework was proposed using advanced retinal image processing, deep learning, and enhanced decision tree algorithm for high-resolution images.

We pre-process retinal image data sets to highlight signs of disease by comparing the patient’s current condition and comparing it to prior images that show the degrees and type of disease using a neural network to extract features of the retinal images, and based on that, the patient’s condition is initially diagnosed.