The evidence for artificial intelligence in eye care has exploded in the past year. Studies show its utility in detecting everything from diabetic retinopathy and age-related macular degeneration to neuroretinal rim damage and even neuropathic corneal nerve pain.1-5 Several research teams have tackled deep learning algorithms for detecting glaucoma, including researchers at Google.6,7 They recently used one approach to predict glaucoma risk and identify optic nerve head (ONH) characteristics—with promising results.7
The team trained their algorithm using 58,033 retinal images already graded by 41 experts and marked as having glaucomatous ONH features, glaucoma risk and gradability. Once trained, the software was testing on 1,205 other images that were also graded by fellowship trained glaucoma specialists. While the system identified glaucoma risk based on the images, it also helped the team evaluate the importance of various ONH features.7
The results, presented last Monday at ARVO 2019, show the algorithm accurately detected the majority of images with risk of glaucoma, with an area under the curve (AUC) of 0.940. The researchers noted the software was more accurate than eye care providers for a group of 411 of the images. The deep learning tool was also accurate when identifying ONH features, with AUCs ranging from 0.608 to 0.977. The team found certain ONH features particularly influenced the glaucoma specialists’ evaluations, including vertical cup-to-disc ratio greater than or equal to 0.7, neuroretinal rim notching, retinal nerve fiber layer defects and bared circumlinear vessels.7
1. De Fauw J, Ledsam JR, Romera-Paredes B, et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nature Medicine. August 13, 2018. [Epub].
2. Rajalakshmi R, Subashini R, Anjana RM, Mohan V. Automated diabetic retinopathy detection in smartphone-based fundus photography using artificial intelligence. Eye (Lond). 2018;32:1138–44.
3. Grassmann F, Mengelkamp J, Brandl C, et al. A deep learning algorithm for prediction of age-related eye disease study severity scale for age-related macular degeneration from color fundus photography. Ophthalmology. 2018 Sep;125(9):1410-20.
4. Thompson AC, Jammal AA, Medeiros FA. A deep learning algorithm to quantify neuroretinal rim loss from optic disc photographs. Am J Ophthalmol. January 26, 2019. [Epub ahead of print].
5. Koseoglu N, Beam A, Hamrah P, et al. The utilization of artificial intelligence for corneal nerve analyses of in vivo confocal microscopy images for the diagnosis of neuropathic corneal pain. Invest Ophthalmol Vis Sci. 2018;59(8):3440.
6. Li Z, He Y, Keel S, et al. Efficacy of a deep learning system for detecting glaucomatous optic neuropathy based on color fundus photographs. Ophthalmology. 2018;125(8):1199-1206.
7. Phene S, Hammel N, Huang AE, et al. Identifying glaucomatous optic nerve head features and glaucoma risk in fundus images at eye-care provider levels of accuracy using deep learning algorithms. ARVO 2019. Abstract 1460 – A0144.