A Deep Learning Based Diabetic Retinopathy Detection from Retinal Images
Mini Yadav, Raghav Goel, D Rajeswari
Abstract
Diabetes is increased tremendously due to metabolism. Lack of early detection, prolonged diabetics might lead to medical complications such as heart problems, eye vision problems, skin issues etc. Diabetic retinopathy (DR) is a frequent abnormality of diabetics. In this paper, we propose computer vision based technique to analyze and predict diabetes from the retinal input images. This helps in an early stage detection of DR. In this image processing steps such as pre-processing, segmentation, feature extraction steps are applied. After the image processing steps, machine learning based classification step is performed. For experimental results, we used python programming language for better results. For experimental results platform, we use jupyter for developing the coding. The framework developed was evaluated on open access public repository datasets, achieving an accuracy of 98.50% using CNN as compared to the accuracy of 87.40% achieved by SVM. These results perform better than several advanced unsupervised ML techniques. It results in decrease of procedural complexity and improved assessment metrics, hence making it suitable to be used in the diagnosis of DR using retinal image analysis.