Enhancing Diabetic Retinopathy Diagnosis Using Intelligent Images Classification

Diabetic is insulin resistance that causes abnormal blood sugar regulation in human beings and damages the kidney, heart, eye, and nerve systems. Retinopathy is the study of retina in organ eyes, which explains the damage that happened to the small blood vessels in the retina. This may cause impaired vision in human. Diabetic patients who are suffering from retinopathy may have blindness. Therefore, it is necessary to detect diabetic retinopathy (DR) stages at an appropriate time. Machine learning algorithms help to categorize diabetic retinopathy stages from retinal images. Thus, the research work on classification helps the doctors to detect diabetic retinopathy. The authors of this article present the novel algorithm Synergistic Image Compression and Feature Extraction (SI COMFEX) for classifying retinal images. Convolutional Neural Network (CNN) and Visual Geometry Group 16 (VGG16) machine learning algorithms or models are configured for multiclass classification. These models successfully classify retinal images into different classes such as mild, moderate, No DR, Proliferative Diabetic Retinopathy (PDR) and severe. The authors of this paper have also compared the designed SI COMFEX algorithm with CNN and VGG16. The performances of the models are verified using Precision, Recall and F1 Score parameters. The comparative study shows that SI COMFEX gives 65.31% testing accuracy, which is better than VGG16 and 10% lower than CNN model.