Faculty Opinions recommendation of Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs.

Publisher Website
Google Scholar
Abstract
Several emerging technologies have demonstrated promise in assisting with the diagnosis of sight threatening retinopathy. This paper describes how researchers in collaboration with Google recently created a deep learning artificial neural network trained to detect retinopathy based on fundoscopic images, and achieved sensitivity of 97.5–96.1% and specificity of 93.4–93.9% in detecting referable disease. In a hypothetical population with a prevalence of 8%, this translates to impressive positive and negative predictive values of 99.8% and of 99.6%, respectively. Such automated systems hold the potential of offsetting the surge in demand for screening. This Recommendation is of an article referenced in an F1000 Faculty Review also written by Matthew Powers, Margaret Greven, Robert Kleinman, Quan Dong Nguyen, and Diana Do.

This publication has 0 references indexed in Scilit: