A Prospective Trial of an Artificial Intelligence based Telemedicine Platform to Stratify Severity of Age-related Macular Degeneration (AMD)

Abstract

Purpose : AMD is a major cause of blindness in developed countries among patients aged 50 and older. Here, we validated an artificial intelligence (AI)-based telemedicine platform for early diagnosis and screening of AMD in a clinical setting.

Methods : We prospectively imaged both eyes of 160 unselected non-dilated subjects aged over 50 years at New York Eye and Ear faculty retina practices with an FDA approved fundus camera (Eidon, Centervue Inc., Fremont ,CA). After initial specialist review, 10 subjects with other confounding conditions like diabetic retinopathy, myopia, vascular occlusion were excluded. 150 eligible subjects (290 eyes, after further exclusion for image quality) were enrolled. All images were uploaded to the telemedicine platform and analyzed by a deep learning algorithm originally developed and tested on the AREDS datasets. (Govindaiah A et al., IEEE EMBC 2018:702-705 ) from iHealthScreen, an independent medical software company. To test the accuracy of the tool, the uploaded images were evaluated by two ophthalmologists and compared against the automated gradings by the software. Patients were classified as referable AMD (intermediate and late AMD) or non-referable (normal macula and early AMD), based on the worst eye. After adjudication of human vs. AI discrepancies in grading (10 human and 7 AI errors), 66 were referable and 84 were non-referable AMD. We considered ‘referable’ as positive and computed the sensitivity and specificity to demonstrate the accuracy of the tool.

Results : For identification of early/none vs. intermediate/late (i.e., referral level) AMD, we achieved 88.67% accuracy with sensitivity of 86.57% and specificity of 90.36%.

Conclusions : A validated color fundus photo-based AI platform for AMD screening is now ready for clinical testing and potential remote telemedical deployment. Creating a more comprehensive, fully effective system also trained on other retinal pathologies for public health service is both warranted and feasible.

This is a 2020 ARVO Annual Meeting abstract.

 

Confusion Matrix on Identification of Referable and Non-referable AMD.