Evaluation of Artificial Intelligence–Based Grading of Diabetic Retinopathy in Primary Care

The results demonstrate both the potential and the challenges of using AI systems to identify diabetic retinopathy in clinical practice. Key challenges include the low incidence rate of disease and the related high false-positive rate as well as poor image quality. Further evaluations of AI systems in primary care are needed.