Artificial intelligence machine learning combined with fundus photography may be able to distinguish between pseudopapilledema and an elevated optic disc when diagnosing optic neuropathies, a South Korean study reports.
The study, published in BMC Ophthalmology, included 295 images of optic neuropathies, 295 images of pseudopapilledema and 779 control images. The researchers defined pseudopapilledema as cases with elevated optic nerve head and blurred disc margin with normal visual acuity (>0.8 Snellen visual acuity), visual field, color vision and pupillary reflex. The optic neuropathy images included cases of ischemic optic neuropathy (177), optic neuritis (48), diabetic optic neuropathy (17), papilledema (22) and retinal disorders (31).
The researchers used four machine-learning classifiers—their investigational model, GoogleNet Inception v3, 19-layer Very Deep Convolution Network (Visual Geometry Group) and 50-layer Deep Residual Learning (ResNet)—to identify the images as normal, pseudopapilledema or papilledema. The investigators noted the networks were trained using approximately 1.2 million images from ImageNet Large-Scale Visual Recognition Challenge.
The researchers reported the accuracy of the machine learning classifiers was very high, ranging from 95.89% to 98.63%. Their model came in at 95.89%, Inception V3 was 96.45%, ResNet was 98.63% and VGG was 96.80%. Additionally, both ResNet and VGG were particularly specific.
|Ahn JM, Kim S, Ahn KS, et al. Accuracy of machine learning for differentiation between optic neuropathies and pseudopapilledema. BMC Ophthalmol. 2019;19(1):178.|