Computer aided diabetic retinopathy detection based on ophthalmic photography: a systematic review and Meta-analysis

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss.
• Accurate detection of DR and its pathological lesions is crucial for timely intervention.
• Computer-aided detection (CAD) systems offer potential for improving diagnostic efficiency.

Purpose of the Study:

• To evaluate the diagnostic value of computer-aided detection (CAD) techniques for diabetic retinopathy (DR) using ophthalmic photography (OP).
• To assess the accuracy of CAD in detecting DR and specific pathological lesions, including exudates (EXs), microaneurysms (MAs), hemorrhages (HMs), and neovascularizations (NVs).

Main Methods:

• A systematic literature search was conducted across multiple databases (PubMed, EMBASE, Ei village, IEEE Xplore, Cochrane Library).
• The Quality Assessment Tool for Diagnostic Accuracy Studies (QUADAS-2) was used to appraise study quality.
• Meta-analysis was performed using Meta-DiSc with a random effects model, and summary receiver operating characteristic (SROC) curves were generated.

Main Results:

• Fourteen articles were included in the meta-analysis.
• Pooled sensitivity and specificity for DR detection by CAD were 90% and 90%, respectively.
• CAD demonstrated high pooled sensitivity and specificity for detecting exudates (89%, 99%), microaneurysms/hemorrhages (42%, 93%), and neovascularizations (94%, 87%).

Conclusions:

• Computer-aided detection (CAD) exhibits high diagnostic accuracy for identifying diabetic retinopathy (DR) and associated pathological lesions from ophthalmic photography.
• Further prospective clinical trials are warranted to validate the efficacy of CAD in DR detection.