A no-code artificial intelligence platform for classifying fluid subtypes in diabetic macular edema: A proof-of-concept workflow using Google Teachable Machine
Hasan Burhanettin Kapti, Erkan DumanAbstract:
This proof-of-concept study evaluated the feasibility of using Google Teachable Machine, a freely available no-code artificial intelligence platform, to classify optical coherence tomography images of diabetic macular edema without programming expertise. A total of 442 OCT images were categorized into intraretinal fluid, combined intraretinal/subretinal fluid, and normal retina. Using default training settings, the model achieved an overall accuracy of 92.4%, with a sensitivity of 91.4%, specificity of 93.2%, and macro F1-score of 0.90. These findings demonstrate that no-code AI platforms may provide an accessible educational and research tool for ophthalmologists, although external validation and larger datasets are required before clinical implementation.