Photo Credit: Graphic_BKK1979

The following is a summary of “Deep learning-based optical coherence tomography and retinal images for detection of diabetic retinopathy: a systematic and meta analysis,” published in the March 2025 issue of Frontiers in Endocrinology by Bi et al. 

Researchers conducted a retrospective study to meta-analyze the effectiveness of deep learning algorithms in detecting diabetic retinopathy (DR) using optical coherence tomography (OCT) and retinal images.  

They searched across multiple databases, including PubMed, Cochrane Library, Web of Science, Embase, and IEEE Xplore, up to July 2024. Studies applying deep learning techniques for detecting DR using OCT and retinal images were included, 2 reviewers independently extracted data and assessed the study quality. A meta-analysis determined pooled sensitivity, specificity, and diagnostic odds ratios.  

The results showed that 47 studies were included in the systematic review, with 10 meta-analyzed, covering 1,88,268 retinal images and OCT scans. The meta-analysis demonstrated a pooled sensitivity of 1.88 (95% CI: 1.45-2.44) and a pooled specificity of 1.33 (95% CI: 0.97-1.84) for DR detection using deep learning models. All deep learning-based OCT outcome odds ratios (ORs) were ≥0.785, indicating that artificial intelligence-assisted methods consistently yielded positive results.  

Investigators concluded that deep learning methods showed high accuracy in identifying DR from OCT and retinal images, suggesting as dependable clinical tools.   

Source: frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2025.1485311/full