Applications of Artificial Intelligence in Optometric Diagnostics: A Review of Techniques and Clinical Impact

This work presents deep learning–based models for automated detection of glaucoma, cataract, and diabetic retinopathy using ophthalmic images. Glaucoma detection utilizes retinal fundus images focusing on the optic disc and cup regions, along with optional optical coherence tomography (OCT) features, to classify normal and diseased cases and assess severity. Cataract detection is performed using slit-lamp and lens images, while diabetic retinopathy is identified from fundus images by detecting features such as microaneurysms, hemorrhages, and exudates. The system incorporates image quality assessment, region of interest extraction, data augmentation, and transfer learning using architectures including ResNet, DenseNet, and MobileNet, along with lightweight CNNs for efficient deployment. Performance is evaluated using accuracy, sensitivity, specificity, F1-score, and AUC, with comparison between segmentation-based and end-to-end approaches. Results demonstrate strong diagnostic performance and highlight model generalizability and computational efficiency for real-world clinical applications.