Alessandro A Jammal1, Atalie C Thompson2, Eduardo B Mariottoni2, Samuel I Berchuck3, Carla N Urata2, Tais Estrela2, Susan M Wakil2, Vital P Costa4, Felipe A Medeiros5. 1. Vision, Imaging and Performance Laboratory (VIP), Duke Eye Center and Department of Ophthalmology, Duke University, Durham, North Carolina, USA; Department of Ophthalmology, State University of Campinas, Campinas, Brazil. 2. Vision, Imaging and Performance Laboratory (VIP), Duke Eye Center and Department of Ophthalmology, Duke University, Durham, North Carolina, USA. 3. Vision, Imaging and Performance Laboratory (VIP), Duke Eye Center and Department of Ophthalmology, Duke University, Durham, North Carolina, USA; Department of Statistical Science and Forge, Duke University, Durham, North Carolina, USA. 4. Department of Ophthalmology, State University of Campinas, Campinas, Brazil. 5. Vision, Imaging and Performance Laboratory (VIP), Duke Eye Center and Department of Ophthalmology, Duke University, Durham, North Carolina, USA. Electronic address: felipe.medeiros@duke.edu.
Abstract
PURPOSE: To compare the diagnostic performance of human gradings vs predictions provided by a machine-to-machine (M2M) deep learning (DL) algorithm trained to quantify retinal nerve fiber layer (RNFL) damage on fundus photographs. DESIGN: Evaluation of a machine learning algorithm. METHODS: An M2M DL algorithm trained with RNFL thickness parameters from spectral-domain optical coherence tomography was applied to a subset of 490 fundus photos of 490 eyes of 370 subjects graded by 2 glaucoma specialists for the probability of glaucomatous optical neuropathy (GON), and estimates of cup-to-disc (C/D) ratios. Spearman correlations with standard automated perimetry (SAP) global indices were compared between the human gradings vs the M2M DL-predicted RNFL thickness values. The area under the receiver operating characteristic curves (AUC) and partial AUC for the region of clinically meaningful specificity (85%-100%) were used to compare the ability of each output to discriminate eyes with repeatable glaucomatous SAP defects vs eyes with normal fields. RESULTS: The M2M DL-predicted RNFL thickness had a significantly stronger absolute correlation with SAP mean deviation (rho=0.54) than the probability of GON given by human graders (rho=0.48; P < .001). The partial AUC for the M2M DL algorithm was significantly higher than that for the probability of GON by human graders (partial AUC = 0.529 vs 0.411, respectively; P = .016). CONCLUSION: An M2M DL algorithm performed as well as, if not better than, human graders at detecting eyes with repeatable glaucomatous visual field loss. This DL algorithm could potentially replace human graders in population screening efforts for glaucoma.
PURPOSE: To compare the diagnostic performance of human gradings vs predictions provided by a machine-to-machine (M2M) deep learning (DL) algorithm trained to quantify retinal nerve fiber layer (RNFL) damage on fundus photographs. DESIGN: Evaluation of a machine learning algorithm. METHODS: An M2M DL algorithm trained with RNFL thickness parameters from spectral-domain optical coherence tomography was applied to a subset of 490 fundus photos of 490 eyes of 370 subjects graded by 2 glaucoma specialists for the probability of glaucomatous optical neuropathy (GON), and estimates of cup-to-disc (C/D) ratios. Spearman correlations with standard automated perimetry (SAP) global indices were compared between the human gradings vs the M2M DL-predicted RNFL thickness values. The area under the receiver operating characteristic curves (AUC) and partial AUC for the region of clinically meaningful specificity (85%-100%) were used to compare the ability of each output to discriminate eyes with repeatable glaucomatous SAP defects vs eyes with normal fields. RESULTS: The M2M DL-predicted RNFL thickness had a significantly stronger absolute correlation with SAP mean deviation (rho=0.54) than the probability of GON given by human graders (rho=0.48; P < .001). The partial AUC for the M2M DL algorithm was significantly higher than that for the probability of GON by human graders (partial AUC = 0.529 vs 0.411, respectively; P = .016). CONCLUSION: An M2M DL algorithm performed as well as, if not better than, human graders at detecting eyes with repeatable glaucomatous visual field loss. This DL algorithm could potentially replace human graders in population screening efforts for glaucoma.
Authors: Mauro T Leite; Linda M Zangwill; Robert N Weinreb; Harsha L Rao; Luciana M Alencar; Pamela A Sample; Felipe A Medeiros Journal: Invest Ophthalmol Vis Sci Date: 2010-03-24 Impact factor: 4.799
Authors: Richard K Parrish; Joyce C Schiffman; William J Feuer; Douglas R Anderson; Donald L Budenz; Maria-Cristina Wells-Albornoz; Ruth Vandenbroucke; Michael A Kass; Mae O Gordon Journal: Am J Ophthalmol Date: 2005-10 Impact factor: 5.258
Authors: Felipe A Medeiros; Pamela A Sample; Linda M Zangwill; Jeffrey M Liebmann; Christopher A Girkin; Robert N Weinreb Journal: Invest Ophthalmol Vis Sci Date: 2006-06 Impact factor: 4.799
Authors: Helen H L Chan; Dai Ni Ong; Yu Xiang G Kong; Evelyn C O'Neill; Surinder S Pandav; Michael A Coote; Jonathan G Crowston Journal: Am J Ophthalmol Date: 2014-02-04 Impact factor: 5.258
Authors: Mauro T Leite; Harsha L Rao; Linda M Zangwill; Robert N Weinreb; Felipe A Medeiros Journal: Ophthalmology Date: 2011-03-05 Impact factor: 12.079
Authors: Tammy M Kuang; Chunwei Zhang; Linda M Zangwill; Robert N Weinreb; Felipe A Medeiros Journal: Ophthalmology Date: 2015-07-18 Impact factor: 12.079
Authors: Evelyn C O'Neill; Lulu U Gurria; Surinder S Pandav; Yu Xiang G Kong; Jessica F Brennan; Jing Xie; Michael A Coote; Jonathan G Crowston Journal: JAMA Ophthalmol Date: 2014-05 Impact factor: 7.389
Authors: Brian C Stagg; Joshua D Stein; Felipe A Medeiros; Barbara Wirostko; Alan Crandall; M Elizabeth Hartnett; Mollie Cummins; Alan Morris; Rachel Hess; Kensaku Kawamoto Journal: Ophthalmol Glaucoma Date: 2020-08-15
Authors: Gala Beykin; Anthony M Norcia; Vivek J Srinivasan; Alfredo Dubra; Jeffrey L Goldberg Journal: Prog Retin Eye Res Date: 2020-07-10 Impact factor: 21.198
Authors: Anna S Mursch-Edlmayr; Wai Siene Ng; Alberto Diniz-Filho; David C Sousa; Louis Arnold; Matthew B Schlenker; Karla Duenas-Angeles; Pearse A Keane; Jonathan G Crowston; Hari Jayaram Journal: Transl Vis Sci Technol Date: 2020-10-15 Impact factor: 3.283