Researchers are finding more ways to apply deep learning technologies to eye care to give optometrists a helping hand in diagnosing disease early and accurately. A recent study shows that artificial intelligence (AI)-based deep learning algorithms were able to differentiate non-glaucomatous from glaucomatous optic neuropathy on color fundus photographs.
In detecting optic disc diseases, the researchers found that the AI’s diagnostic accuracy to distinguish glaucomatous optic neuropathy from non-glaucomatous images was substantial, with a sensitivity of 93.4% and specificity of 81.8%.
Researchers used 3,815 fundus images consisting of 2,883 normal optic disc images, 446 images of eyes with non-glaucomatous optic neuropathy with optic disc pallor and 486 images of glaucomatous optic neuropathy. The images were evaluated by the AI and by two expert neuro-ophthalmologists. All of the glaucoma images had corroborating evidence of disease on visual field testing and optical coherence tomography.
The researchers did note that false positive cases were found in images with extensive areas of peripapillary atrophy and tilted optic discs.
|Yang H, Kim YC, Sung J, et al. Efficacy for differentiating nonglaucomatous versus glaucomatous optic neuropathy using deep learning systems. Am J Ophthalmol. April 2, 2020. [Epub ahead of print].|