A review on computer-aided recent developments for automatic detection of diabetic retinopathy

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss, characterized by microvascular damage in the retina.
• Manual detection of DR features like microaneurysms, exudates, and hemorrhages from fundus images is labor-intensive and prone to error.
• Early and automated detection of DR is crucial for timely intervention and preventing blindness.

Purpose of the Study:

• To systematically review and analyze various automated methods for diabetic retinopathy detection and classification.
• To discuss retinal blood vessel detection techniques relevant to diagnosing proliferative diabetic retinopathy.
• To provide a comprehensive overview of feature extraction, segmentation, and classification techniques used in DR diagnosis systems.

Main Methods:

• Systematic literature review of publicly available databases and medical sources.
• Meta-analysis of diverse methods for diabetic retinopathy feature extraction and segmentation.
• Evaluation of various classifiers for system performance metrics in DR diagnosis.

Main Results:

• The review covers multiple automated approaches for DR detection and grading based on severity.
• Retinal blood vessel detection methods are discussed for improved diagnostic accuracy.
• Performance metrics of different systems are analyzed through meta-analysis.

Conclusions:

• Automated systems offer a more efficient and potentially more accurate alternative to manual DR screening.
• This survey highlights key methods and challenges, guiding future research in enhancing DR diagnostic systems.
• The findings are valuable for researchers and technical professionals aiming to improve real-world DR diagnosis.