1. A survey on recent developments in diabetic retinopathy detection through integration of deep learning.
- Author
-
Agarwal, Shalini and Bhat, Aruna
- Subjects
DEEP learning ,DIABETIC retinopathy ,HYPERGLYCEMIA ,GENERATIVE adversarial networks ,COMPUTER-aided diagnosis ,BLOOD sugar - Abstract
Diabetes, nowadays, is a very common disease throughout the world among people of all ages. The higher level of blood sugar in the blood more often leads to Damage to blood vessels in the retina leads to blindness if left untreated or undetected on time. Diabetic Retinopathy (DR) screening programs like retinal fundus image analysis help ophthalmologists deal with some visual impairment problems. Computer-Aided Diagnosis aims to detect the severity of DR as early as possible so that it can be handled before the occurrence of any irreversible vision loss. With the help of many advancements in artificial intelligence techniques, a highly efficient and accurate system can be designed to help medical professionals automatically diagnose DR at an early stage without any special clinical resources. This paper conducts a thorough investigation into several recent frameworks proposed based on machine learning and deep learning networks to classify non-proliferative diabetic retinopathy, exudates, hemorrhages, and micro aneurysms. Several promising pre-trained deep learning model to classify DR stages exploited by researchers, and also investigated Transfer learning on pre-trained GoogLeNet and AlexNet, VGG etc. models. It is observed that almost all public and private data sets widely available for research are imbalanced. To alleviate these issues, generative adversarial networks (GANs) and their variants were also used to generate label-preserving data. In this study, the authors also list the recently proposed GAN-based frameworks and their impact on model performance. The paper concludes with the current challenges and future directions in the early and accurate classification of the severity level of DR. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF