A Comprehensive Review of Diabetic Retinopathy Detection and Grading Based on Deep Learning and Metaheuristic Optimization Techniques

Diabetic retinopathy (DR) is a microvascular disorder that causes retinal damage and irreversible blindness. It is a condition instigated by prolonged Diabetes Mellitus. According to statistics, one-third of people with Diabetes are likely to have DR-induced vision impairment. DR diagnosis is crucia...

Full description

Saved in:
Bibliographic Details
Published in:Archives of computational methods in engineering 2023-09, Vol.30 (7), p.4565-4599
Main Authors: Dayana, A. Mary, Emmanuel, W. R. Sam
Format: Article
Language:English
Subjects:
Algorithms
Blindness
Classification
Datasets
Deep learning
Diabetes
Diabetes mellitus
Diabetic retinopathy
Engineering
Heuristic methods
Machine learning
Mathematical and Computational Engineering
Medical imaging
Optimization
Optimization techniques
Physical work
Review Article
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Diabetic retinopathy (DR) is a microvascular disorder that causes retinal damage and irreversible blindness. It is a condition instigated by prolonged Diabetes Mellitus. According to statistics, one-third of people with Diabetes are likely to have DR-induced vision impairment. DR diagnosis is crucial at the primary stages, as delayed treatment can impair vision and increase the risk of blindness. Manually scrutinizing the fundus images is tedious because it takes considerable time, effort, and skill to identify the severity level of retinopathy. However, the recent advancements in deep learning have unlocked a new arena for researchers. An effective deep learning-based screening method can improve health care by minimizing manual labor and exploiting limited resources. This review article comprehensively analyses the deep learning techniques used in diabetic retinopathy detection and grading based on fundus images. The available fundus image datasets and the preprocessing methods employed for DR classification have been analyzed. Further, DR classification based on transfer learning, ensemble learning, and metaheuristic optimization algorithms is systematically reviewed and discussed. Finally, the review summarizes the challenges and limitations of the published DR grading algorithms and suggests future directions that could bring effective solutions. The comparative analysis of the reviewed studies shows that most deep learning methods outperform the conventional methods achieving 99% accuracy on benchmark datasets.
ISSN:1134-3060
1886-1784
DOI:10.1007/s11831-023-09946-5