Diabetic retinopathy detection through artificial intelligent techniques: a review and open issues

Diabetic Retinopathy (DR) is the disease caused by uncontrolled diabetes that may lead to blindness among the patients. Due to the advancements in artificial intelligence, early detection of DR through an automated system is more beneficial over the manual detection. At present, there are several published studies on automated DR detection systems through machine learning or deep learning approaches. This study presents a review on DR detection techniques from five different aspects namely, datasets, image preprocessing techniques, machine learning-based approaches, deep learning-based approaches, and performance measures. Moreover, it also presents the authors’ observation and significance of the review findings. Furthermore, we also discuss nine new research challenges in DR detection. After a rigorous selection process, 74 primary publications were selected from eight academic databases for this review. From the selected studies, it was observed that many public datasets are available in the field of DR detection. In image preprocessing techniques, contrast enhancement combined with green channel extraction contributed the most in classification accuracy. In features, shape-based, texture-based and statistical features were reported as the most discriminative in DR detection. The Artificial Neural Network was proven eminent classifier compared to other machine learning classifiers. In deep learning, Convolutional Neural Network outperformed compared to other deep learning networks. Finally, to measure the classification performance, accuracy, sensitivity, and specificity metrics were mostly employed. This review presents a comprehensive summary of DR detection techniques and will be proven useful for the community of scientists working in the field of automated DR detection techniques.