Automated Screening for Diabetic Retinopathy - A Systematic Review

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetes prevalence is rising globally, increasing the burden of diabetic retinopathy (DR).
• Diabetic retinopathy can lead to severe vision impairment and blindness.
• Current DR screening methods are costly and labor-intensive, necessitating efficient alternatives.

Purpose of the Study:

• To systematically review studies on automated retinal image analysis for diabetic retinopathy detection.
• To assess the performance of automated systems in simulated or actual clinical screening settings.
• To evaluate the potential of automated analysis to improve DR screening efficiency.

Main Methods:

• A systematic review was conducted using PubMed, Cochrane Library, and Embase.
• Searches identified 1,231 publications, with further manual searches performed.
• Four screening levels resulted in the inclusion of 7 relevant studies.

Main Results:

• Seven studies were included in the review.
• Automated DR detection demonstrated high sensitivity (87.0-95.2%) but lower specificity (49.6-68.8%).
• False negatives often involved mild DR; diabetic macular edema was missed in some studies. Meta-analysis was not feasible due to study heterogeneity.

Conclusions:

• Automated retinal image analysis shows promise for DR screening due to high sensitivity.
• Despite limitations in specificity, the technology offers significant workload reduction.
• Automated systems may be valuable in various diabetic retinopathy screening contexts.