Fundus Imaging

Fundus imaging studies how photographs of the retina can reveal diseases such as diabetic retinopathy, glaucoma, and age-related macular degeneration, as well as markers of broader health. Common approaches use CNNs and vision transformers for disease grading, U-Net variants for segmenting vessels and optic-disc structures, and foundation or vision-language models for broader diagnostic tasks. Researchers also use anatomical priors, multimodal data such as OCT, and explanation methods to make predictions more informative and clinically useful; datasets such as DRIVE, STARE, and EyePACS support evaluation across tasks. A central challenge is reliable performance across cameras, populations, and disease stages: large pretrained models do not always outperform smaller CNNs, while inconsistent labels and limited external validation make clinical generalization difficult to establish.