Moore’s Field Eye Hospital partnered with Google’s DeepMind to develop OCT algorithms for the eye. They can diagnose most retinal disease: diabetic retinopathy, AMD, glaucoma, and fifty more with a combination of high-quality photos of the eye and deep learning algorithm. Smart phone attachments can enable remote OCT or fundoscopic photo, coupled with an algorithm to diagnose emerging eye conditions. Also, AI algorithms can assess risk of cardiovascular disease by looking at the same funduscopic photos and examining the vessels for evidence of vascular disease. This can give window to the body and in the future, periodic use of smart phone eye exams for monitoring health may become part of the ongoing screening and management of chronic conditions.
AI models have shown promise in predicting the potential progression of early Age-related Macular Degeneration (AMD) to clinically significant disease. While millions of individuals show early age-related macular degeneration (AMD) signs, yet have excellent vision, the risk of progression to advanced AMD with legal blindness is highly variable. Artificial intelligence with automated analysis of imaging biomarkers allows personalized prediction of AMD progression. Deep learning-based automated screening method for finding individuals at risk of developing clinically significant Age-related Macular Degeneration (AMD) can help with this process. A color fundus image allows for categorization of AMD to none/early AMD or intermediate/ advanced comparable to expert Ophthalmologist accuracy.
DeepSeeNet, a deep learning model, was developed to classify patients with AMD based on the severity of their disease . DeepSeeNet was trained on 58,402 and tested on 900 images from the longitudinal follow-up of 4,549 participants. DeepSeeNet performed better on patient-based classification than retinal specialists with high AUC in the detection of large drusen (0.94), pigmentary abnormalities (0.93), and late AMD (0.97). DeepSeeNet also outperformed retinal specialists in the detection of large drusen and pigmentary abnormalities but showed lower performance in the detection of late AMD (accuracy 0.967 vs. 0.973; kappa 0.663 vs. 0.754). By simulating the human grading process, DeepSeeNet demonstrated high accuracy with increased transparency in the automated assignment of individual patients to AMD risk categories. These results highlight the potential of deep learning to assist and enhance clinical decision-making in patients with AMD, such as early AMD detection and risk prediction for developing late AMD.
