Diabetic Retinopathy Detection using Deep Learning: A Review

Abstract

Diabetic Retinopathy is a diabetes-related complication that affects the blood vessels in the retina of the eye. It is a leading cause of vision impairment and blindness worldwide if not treated on time. Early detection of this disease is crucial for timely intervention and prevention of vision loss. The ophthalmologist manually examines the morphological changes in retinal veins and lesions in fundus images. This process is time-consuming, expensive, and complex. Thus procedure can be streamlined with the help of computer-aided diagnostic systems (CADs) used for identifying DR lesions. Artificial Intelligence and Deep learning algorithms offer a more efficient and objective approach to analyzing retinal images, enabling automated detection of distinctive features associated with diabetic retinopathy. As a review, this paper concentrates on applications of deep learning models and transfer learning in diabetic retinopathy detection. Many publications are explored and various methods for preprocessing, segmentation, class balancing, feature extraction, data augmentation and classification used in these publication are discussed. This paper also provides an overview of performance results achieved by these state-of-the-art methodologies in published papers. The review highlights benefits and challenges of current approaches, offering valuable insights for other researchers.