Pegsys

Glaucoma, a leading cause of irreversible blindness, remains challenging to diagnose due to its asymptomatic progression. Traditional diagnostic methods require high-resolution fundus imaging and specialized expertise, limiting their feasibility for widespread screening, particularly in resource-limited areas, such as rural regions and remote mountainous areas in China. To address this issue, we propose U-Net-Segformer, a glaucoma screening framework designed to enhance segmentation accuracy for low-quality images and mitigate performance degradation across datasets. The model achieved Dice coefficients of 91.01% and 97.50% for optic cup (OC) segmentation and 89.39% and 96.50% for optic disc (OD) segmentation on the Drishti-GS and REFUGE datasets, with absolute cup-to-disc ratio errors of 3.13% and 3.76%, respectively. By integrating U-Net-Segformer with portable fundus imaging devices and cloud-based AI analysis, we developed an intelligent, mobile glaucoma screening system capable of real-time diagnosis. The system features a user-friendly mobile terminal that enables fundus image capture anywhere, with cloud-based processing for real-time screening results. This approach addresses the shortage of glaucoma screening resources in primary care and remote areas. Extensive evaluations on public and self-collected datasets from partner hospitals demonstrated a 94% accuracy rate, confirming the system’s effectiveness and clinical applicability

Economic shortage · Fundus (uterus) · Glaucoma · Image segmentation · Limiting · Optic disc · Segmentation · AI in cancer detection · Glaucoma and retinal disorders · Retinal Imaging and Analysis