An artificial intelligence (AI) system is now capable of identifying vision-threatening conditions in high myopic patients. The global myopia epidemic has made this type of technology more important than ever, as myopia is a risk factor for a number of conditions that will continue to increase the burden on healthcare systems in the next few decades.
Researchers from China conducted a cross-sectional, prospective study and obtained 5,505 OCT macular images from 1,048 patients with high myopia to develop the AI system. The researchers trained four independent convolutional neural network models on these images to identify retinoschisis, macular hole, retinal detachment and pathological myopic choroidal neovascularization.
The independent test dataset included 412 images from 91 high myopes. In this set, areas under the receiver operating characteristic curves were high for all conditions (0.961 to 0.999). The researchers reported that the AI system achieved greater than 90% specificity, and its sensitivity was equal to or better than those of retina specialists. “Our AI system provided a transparent and interpretable diagnosis with heatmaps,” they wrote in their paper.
The researchers concluded that their model may be used for large-scale myopia screening and patient follow-up.
Li Y, Feng W, Zhao X, et al. Development and validation of a deep learning system to screen vision-threatening conditions in high myopia using optical coherence tomography images. Br J Ophthalmol. December 21, 2020. [Epub ahead of print].