High Confidence Artificial Intelligence (AI) Predictions in Glaucoma Detection: A RIM ONE Database Study

Introduction and Objectives: Glaucoma is a progressive optic neuropathy that can lead to irreversible blindness. This study evaluates the use of neural networks in glaucoma prediction with high confidence. Patients or Materials and Methods: The RIM One dataset was used, training an EfficientNetV2B0 model on fundus images. A 95% threshold was set for high-confidence predictions. Results: Sensitivity was 91% and specificity was 99%. Applying the highconfidence threshold increased the AUC to 100%. Conclusions: This study demonstrates the feasibility of using highconfidence AI predictions for glaucoma diagnosis, improving clinical relevance.