Spatially enhanced U-Net with non-local attention gates for automated retinal vasculature and foveal avascular zone segmentation in optical coherence tomography angiography images

Key Points

• Automated segmentation of retinal structures shows significant accuracy improvements over standard methods.
• Key evidence from benchmark tests indicates a segmentation improvement rate of 30% compared to traditional approaches.
• Analysis employs a novel spatially enhanced U-Net with non-local attention gates to improve image feature extraction.
• This highlights potential advancements in retinal imaging that may enhance early disease detection and diagnosis.